Data Processing Agreement
This Data Processing Agreement (DPA) sets out how we handle personal data you entrust to us when you use Bidwhistle. It is written to meet Article 28 of the GDPR and the equivalent UK GDPR requirements, and it sets out the transfer clauses that cover the onward transfers to our providers — section 13 explains which legs of the journey need a mechanism and which are covered by adequacy. If your legal or procurement team needs a signed copy, email legal@bidwhistle.com.
1. Definitions
1.1 Capitalised words not defined here have the meaning given in the Terms of Service — in particular Agreement, Service, AI Features, Olivia, Input, Output, Customer Data, Procurement Data, Usage Data, Authorised User and Subscription Period.
1.2 In this DPA:
“Customer Personal Data” means the Personal Data contained in Customer Data that we Process on your behalf under the Agreement, including Inputs and Outputs to the extent they contain Personal Data. It does not include Procurement Data, Usage Data, or your account and billing data, for which we act as an independent Controller (section 3).
“Data Protection Law” means all law applicable to Processing under this DPA, including the GDPR, the UK GDPR, the Estonian Personal Data Protection Act (isikuandmete kaitse seadus), the ePrivacy Directive 2002/58/EC and PECR, the Swiss Federal Act on Data Protection (FADP), and any successor legislation.
“DPF” means the EU–US Data Privacy Framework, together with the UK Extension to it and the Swiss–US Data Privacy Framework.
“GDPR” means Regulation (EU) 2016/679. “UK GDPR” means the UK version, read with the Data Protection Act 2018 as amended by the Data (Use and Access) Act 2025.
“Personal Data”, “Controller”, “Processor”, “Data Subject”, “Processing” and “Supervisory Authority” have the meanings given in the GDPR.
“Restricted Transfer” means a transfer or onward transfer of Personal Data subject to Chapter V of the GDPR, Chapter V of the UK GDPR or the equivalent provisions of the FADP, so that a transfer mechanism is required.
“SCCs” means the standard contractual clauses annexed to Commission Implementing Decision (EU) 2021/914 of 4 June 2021, as amended or replaced.
“Security Incident” means a breach of security leading to the accidental or unlawful destruction, loss, alteration, unauthorised disclosure of, or access to, Customer Personal Data. It does not include unsuccessful attempts that do not compromise security — blocked logins, port scans, failed authentication, pings and similar.
“Sub-processor” means a third party we engage to Process Customer Personal Data on our behalf, including our infrastructure, AI and business-operations providers.
“UK Addendum” means the International Data Transfer Addendum to the EU Commission Standard Contractual Clauses, version B.1.0, issued by the Information Commissioner under section 119A of the Data Protection Act 2018 and laid before Parliament on 2 February 2022.
2. Applicability, scope and order of precedence
2.1 This DPA forms part of the Agreement. It is incorporated into the Terms of Service by reference and takes effect when you accept the Terms of Service or first use the Service, whichever is first. No separate signature is needed for it to bind us both.
2.2 If you need an executed copy, email legal@bidwhistle.com and we will sign the version published at Data Processing Agreement as at the date of your request. We do not charge for this, and signing does not change the terms.
2.3 Where you contract for a group, you enter into this DPA for yourself and as agent for each affiliate using the Service under your account, and you remain responsible for their compliance.
2.4 This DPA applies to our Processing of Customer Personal Data on your behalf, for as long as we Process it. It does not apply where we act as an independent Controller (sections 3.4 and 3.5); that Processing is described in our Privacy Notice.
2.5 This DPA prevails over the rest of the Terms of Service on data protection matters. Everything else in the Terms of Service continues to apply. If there is a conflict, this order applies, highest first:
| Rank | Document |
|---|---|
| 1 | The SCCs and, for UK transfers, the UK Addendum |
| 2 | The body of this DPA |
| 3 | Your Order or plan details |
| 4 | The Terms of Service |
| 5 | The Acceptable Use Policy, the Service Level Agreement and the Refund and Cancellation Policy |
This order governs data protection matters only; for everything else the order of precedence in the Terms of Service applies.
2.6 Nothing here is intended to conflict with the SCCs or reduce the protection they give. If it does, the SCCs prevail.
3. Roles of the parties
3.1 We handle three categories of data in three different capacities.
| Data | Your role | Our role | In one sentence |
|---|---|---|---|
| Customer Data (including Inputs, Outputs and documents you upload to the Response Workspace) | Controller — or Processor, where you process on behalf of another controller | Processor — or Sub-processor, where you are a Processor | You decide what you put into Bidwhistle and why; we handle it only to run the Service for you and on your instructions. |
| Account, billing and Usage Data | — | Independent Controller | We decide how to bill you, secure the platform and improve the product, so for that data we answer to Data Subjects directly. |
| Procurement Data (including named public-sector contacts) | — | Independent Controller | We collect public procurement information from official sources on our own account, on the basis of our legitimate interests and with the transparency information Article 14 GDPR requires. |
3.2 For Customer Personal Data you are the Controller and we are the Processor. We Process it only as set out in section 4.
3.3 Where you are a Processor: (a) you confirm you have your controller’s authorisation to appoint us and to agree this DPA, including the SCCs, on their behalf; (b) Module Three of the SCCs applies to any Restricted Transfer instead of Module Two; and (c) references to your instructions mean instructions you are permitted to give under your own contract with that controller, and we may rely on them.
3.4 Account, billing and Usage Data. We are an independent Controller of your account and profile details, subscription and payment records, support correspondence, and technical and usage information. Our Privacy Notice sets out the lawful bases and retention periods.
3.5 Procurement Data. We are an independent Controller of Procurement Data, including the name, job title, employer, work email and work telephone number of a public-sector contact as published in a notice. Procurement Data is not Customer Data and is not Processed on your behalf. Our lawful basis is legitimate interests under Article 6(1)(f) GDPR; we keep a Legitimate Interests Assessment on file and publish Article 14 transparency information, including a no-account-needed route to object, in our Privacy Notice.
3.6 Nothing here makes us joint controllers under Article 26 GDPR. Each of us is separately responsible for our own compliance in the capacity above.
3.7 Documents you upload to the Response Workspace. The Response Workspace lets you upload a tender pack and have it read, have answers drafted, have a requirement explained and have a draft bid reviewed.
- You are the Controller of all Personal Data contained in a document you upload — not only personal data about your own staff, but personal data about anyone else named in it. Tender packs routinely carry personal data about people who have no relationship with you or with us: most often a TUPE schedule naming the incumbent supplier’s employees with their roles, salaries and length of service.
- We are the Processor of that Personal Data. We process it only to answer the request you have made, we pass it to the AI Sub-processor identified in Annex VI for that purpose, and we do not store the pack (Annex I(B), section 6).
- Uploading is a disclosure. When you upload a document you disclose the Personal Data in it to us and, through us, to that Sub-processor. Making that disclosure lawfully is your responsibility, not ours: we do not see the document before you send it and we are not in a position to judge whether you were entitled to share it.
- We apply three controls at the point of upload: a notice, before upload, explaining that packs for staffed contracts usually include a schedule of the people currently doing the work and that we leave those files out; a list of every file in the pack, of which only the files you leave ticked are read and transmitted to our Sub-processor; and the listing of files whose names indicate a staff record as unticked by default. Three limits apply. The default is not a block: you may overrule it, and a file you tick is transmitted. The check operates on file names, not file contents, so a staff schedule whose file name does not disclose its nature is not recognised and is presented ticked. And it operates on whole files, so Personal Data inside a file you tick is transmitted with it and is not separated out. These controls reduce the likelihood of an inadvertent disclosure; they do not prevent one, and deciding what to include remains your responsibility under section 4.8.
4. Processing instructions and purpose limitation
4.1 We Process Customer Personal Data only on your documented instructions, including on transfers to a third country, unless required to do otherwise by Union or Member State law we are subject to.
4.2 Your documented instructions are: (a) this DPA, including Annex I(B); (b) the rest of the Agreement; (c) your and your Authorised Users’ configuration and use of the Service — creating a saved search, asking Olivia a question, uploading a document to the Response Workspace and asking us to work on it; and (d) any further written instruction we agree with you. Together these are your complete and final instructions at the effective date.
4.3 If you need Processing outside those instructions we will discuss it. We may decline where it is not technically feasible or would require us to change the Service, and we may charge a reasonable fee for work materially beyond the Service.
4.4 Instructions that infringe the law. We will tell you promptly if, in our opinion, an instruction infringes Data Protection Law. We may suspend the Processing concerned until the instruction is confirmed, corrected or withdrawn, and we are not in breach for doing so. We are not obliged to give you a legal opinion on your instructions.
4.5 We do not use Customer Personal Data for our own purposes. We do not sell it, share it with third parties for their own purposes, use it for advertising, or use it to build or improve products for anyone but you. The only exceptions are:
- providing, maintaining, securing and supporting the Service for you;
- complying with a legal obligation (section 4.6);
- creating aggregated and de-identified insights and benchmarks as permitted by the Terms of Service — they must not identify you, any Authorised User or any individual, and we covenant not to attempt to re-identify them or attribute them to an identifiable customer; and
- enforcing the Acceptable Use Policy or investigating a suspected Security Incident, where access is least-privilege, minimal and logged.
4.6 If law requires us to Process other than on your instructions, we will tell you before Processing unless the law prohibits it on important grounds of public interest. Section 13.8 also applies to requests from public authorities.
4.7 Your responsibilities. You are responsible for the accuracy, quality and legality of Customer Personal Data and how you obtained it; for having a lawful basis and giving Data Subjects the required transparency information; for not submitting data you are not entitled to submit or the special categories described in Annex I(B); for securing your credentials and managing your Authorised Users; and for satisfying yourself that the measures in Annex II suit the risk of the Processing you instruct.
4.8 Your warranty on documents you upload. Each time you or an Authorised User uploads a document to the Response Workspace, you warrant that:
- you have a lawful basis under Article 6 GDPR for the Personal Data in it, including Personal Data about people who are not your own staff, and you have given or are exempt from giving those people the transparency information Articles 13 and 14 GDPR require;
- you are entitled to disclose that Personal Data to us and, through us, to the AI Sub-processor identified in Annex VI, and doing so does not breach a duty of confidence, a contract, a procurement rule or any other obligation you are under;
- you have removed or redacted Personal Data we do not need in order to answer your request — a tender pack usually needs far less personal data than it contains; and
- the document contains no special category data under Article 9 GDPR and no criminal offence data under Article 10 GDPR.
4.9 This warranty is given each time you upload, and it is backed by the customer indemnity in the Terms of Service. We do not verify it, and nothing in this DPA obliges us to. If you tell us a document should not have been uploaded, tell us at privacy@bidwhistle.com and we will help you deal with the consequences under section 9.
5. AI training
This section is drafted to be precise rather than reassuring. It separates what we commit to, which is unqualified, from what each AI Sub-processor has confirmed, which varies — and it says plainly where a confirmation is outstanding.
5.1 Our own commitment
5.1.1 We do not use Customer Data, Inputs or Outputs to train AI models. We do not train, fine-tune, retrain, align, distil or otherwise develop or improve any generative AI or machine learning model of our own using Customer Personal Data, Inputs or Outputs. We do not put them in evaluation or benchmark datasets, and we do not submit them for human review for model improvement.
5.1.2 We do not sell, licence or share Customer Personal Data, Inputs or Outputs to or with any third party so that the third party can train on them.
5.1.3 These commitments are contractual, they apply for the whole term of the Agreement, and they do not depend on any provider’s cooperation. They are ours to keep and we keep them.
5.2 What our AI sub-processors have confirmed
5.2.1 We cannot honestly extend 5.1 to every provider in the chain, so we set out the position provider by provider. Annex VI and our Sub-processors page carry the same positions, and are updated together.
| AI Sub-processor | Training position, as at the date of this DPA |
|---|---|
| Anam — Olivia’s avatar, voice, speech-to-text and hosted reasoning | Confirmed no training. Anam has confirmed that it does not use conversations for model training |
| Google — the reasoning model behind Olivia, contracted through Anam | No specific commitment. The model is reached through Anam and runs under Google’s standard terms, which do not give us a no-training commitment. This is the weakest data commitment in our chain |
| Anthropic — the reasoning model behind the Response Workspace | No — Anthropic's Commercial Terms state: “Anthropic may not train models on Customer Content from Services.” |
| Voyage AI — semantic retrieval for the knowledge base | No — Bidwhistle has exercised Voyage's opt-out (10 September 2026). ⚠️ Voyage's terms take a training licence BY DEFAULT unless the customer opts out, so this rests on our opt-out rather than on their default position. |
5.2.2 We will update this section, Annex VI and the Sub-processors page when we have the answers, and we will notify subscribers to sub-processor notifications when a position changes. We hold no written confirmation of training, retention or processing location for the Google model reached through Anam, and none of the processing location for Voyage AI. We are seeking both and will record them here. Anthropic’s position is taken from its published terms.
5.2.3 We have not written “no training” against a provider that has not told us so. If a column in Annex VI says a position is unconfirmed, that is the fact, not a drafting omission.
5.3 What we do about it
5.3.1 For each AI Sub-processor we: (a) ask for, and where we can obtain it contract for, a prohibition on using our customers’ data for model training; (b) select the zero-retention or minimum-retention setting where the provider makes one available to us on our plan — see 5.3.2; (c) turn off any feedback, rating or telemetry feature that would send Inputs or Outputs to the provider for its own purposes; and (d) record the provider’s retention and training position in Annex VI and on our Sub-processors page, so you can check rather than take our word for it.
5.3.2 An honest limit on 5.3.1(b). Our voice provider offers both regional processing and zero data retention, but neither is available on the plan we currently hold — they are account entitlements rather than settings, and a request for either is refused at the provider’s end. The consequence is set out in section 13 and Annex VI: Olivia’s session recording and transcript are held by that provider for 30 days and the processing region is the provider’s choice and may be the United States. We are working to move to a plan that lets us pin the region to the EU and switch retention off. Region selection and zero data retention are not available to us on the plan we hold. We are discussing both with the provider.
5.3.3 We may use aggregated technical statistics containing no Personal Data and no customer content — request counts, token counts, latency, error rates, feature usage — to operate and improve the Service.
5.3.4 We do not routinely read the content of your Olivia conversations, your uploaded documents, your saved searches or your notes. Where a person needs access — you have asked for support, we are investigating a suspected Security Incident or a report under the Acceptable Use Policy, or the law requires it — access is minimal, least-privilege and logged.
5.3.5 If we intend to change the commitment in 5.1 we will give at least 30 days’ notice before it takes effect, and you may terminate the affected part of the Service and receive a pro-rata refund for the unused remainder of your Subscription Period. We have no plan to change it. A change in a provider’s position is notified under section 8 and Annex VI, and carries the same objection right.
6. Confidentiality of personnel
6.1 Access to Customer Personal Data is limited to people who need it to perform our obligations. Bidwhistle is operated by a very small team, so in practice that is a small number of named individuals.
6.2 Everyone we allow to Process Customer Personal Data — employees, contractors and anyone else acting for us — is bound by a written duty of confidentiality or an appropriate statutory obligation of confidentiality, which survives the end of their engagement.
6.3 Everyone with access receives data protection and security guidance appropriate to their role before access is granted. We review access at least every six months and remove what is no longer needed.
6.4 We are responsible for the acts and omissions of our personnel in relation to Customer Personal Data as if they were our own.
7. Security
7.1 We implement and maintain appropriate technical and organisational measures to ensure a level of security appropriate to the risk, as Article 32 GDPR requires. They are described in Annex II.
7.2 In assessing what is appropriate we take account of the state of the art, the costs of implementation, the nature, scope, context and purposes of the Processing, and the risk to the rights and freedoms of Data Subjects.
7.3 Updates. We may update the measures in Annex II from time to time, provided the updates do not materially reduce the overall level of security of the Service. The current version is always published with this DPA.
7.4 Security is shared. You are responsible for configuring the Service appropriately, using the security features we make available (including multi-factor authentication where offered), managing your Authorised Users and keeping credentials confidential.
7.5 We hold no security certification — no ISO 27001, no SOC 2, no Cyber Essentials — and we do not claim one. Annex II describes what we actually do and marks what is planned rather than in place.
8. Sub-processors
8.1 You give us a general written authorisation under Article 28(2) GDPR to engage Sub-processors, subject to this section 8. Those authorised at the effective date are listed at Sub-processors and reproduced in Annex III. The published page is the authoritative and current list.
8.2 Notice. We will give at least 30 days’ notice before a new Sub-processor starts Processing Customer Personal Data, or before an existing one takes on a materially different role. We give notice by updating the Sub-processors page and emailing everyone subscribed to sub-processor notifications; you can subscribe from that page. Please make sure someone at your organisation is subscribed — it is how we reach the right person.
8.3 An honest note about AI providers. Some AI and inference providers give us less than 30 days’ notice of their own changes, and occasionally we must change a model provider quickly to keep Olivia working. Where that happens we cannot give 30 days’ advance notice. We will then notify you as soon as reasonably practicable, and no later than the date the change takes effect. Your objection right below applies in full, and the 30-day objection period runs from the date of our notice rather than the date of the change.
8.4 Objection. You may object to a new Sub-processor within 30 days of our notice, on reasonable grounds relating to data protection, by writing to privacy@bidwhistle.com and setting out the grounds. If you do not object within that period you are deemed to have accepted the Sub-processor.
8.5 If you object, we will work with you in good faith for up to 30 days to find a reasonable resolution — explaining the safeguards, changing how the Sub-processor is configured, excluding your data where technically possible, or offering an alternative.
8.6 If we cannot resolve your objection in that period, you may terminate the affected part of the Service on written notice (or the Agreement, if that part is essential to it) and we will refund the fees paid for that part on a pro-rata basis for the unused remainder of your Subscription Period. That is your exclusive remedy for an unresolved objection, and neither of us has further liability for the termination itself.
8.7 Flow-down. We appoint each Sub-processor under a written contract imposing data protection obligations no less protective than this DPA, as Article 28(4) GDPR requires — on confidentiality, security, sub-processing, assistance, breach notification, deletion and international transfers. Where a Sub-processor is outside the EEA and the United Kingdom, a transfer mechanism under section 13 applies to it.
Two qualifications, because a flow-down claim is only worth what it can be evidenced by. First, for the two AI providers we contract with directly, the position is now taken from their published terms rather than from confirmations we were waiting on. The reasoning model behind the Response Workspace (Anthropic) commits in its Commercial Terms that it “may not train models on Customer Content from Services”, and its Data Processing Addendum, carrying the EU Standard Contractual Clauses, is incorporated automatically when those terms are accepted. The semantic retrieval provider (Voyage AI) licenses Customer Content for training BY DEFAULT unless the customer opts out — and Bidwhistle has exercised that opt-out (10 September 2026), which also gives zero-day retention on its hosted endpoints. Anthropic now publishes that it stores data in the United States and deletes API inputs and outputs within 30 days, with stated exceptions; Voyage AI publishes no processing location. Annex III and Annex VI record both positions. Second, two providers in the chain — the model behind Olivia, and Cartesia, which generates her voice — are reached through another provider rather than contracted by us, so the terms covering them are that provider’s (section 13.7.2).
8.8 We remain fully liable to you for each Sub-processor’s performance of its data protection obligations, subject to section 15. On written request we will provide a copy of the data protection terms agreed with a Sub-processor, redacted to remove commercial information not relevant to data protection.
9. Assistance to the customer
9.1 Data subject requests. The Service includes tools to access, correct and export Customer Personal Data yourself, and to ask for an account to be deleted; those are usually the fastest route. Deletion is carried out by us within the periods in section 11 rather than taking effect the moment it is asked for. Where they are not enough, we will provide reasonable assistance, taking into account the nature of the Processing and the information available to us, to help you respond to requests under Chapter III GDPR — access, rectification, erasure, restriction, portability, objection, and rights relating to automated decision-making.
9.2 If a Data Subject contacts us directly about Customer Personal Data, we will not respond on the substance unless you tell us to or the law requires it. We will tell you without undue delay and, where we can identify the customer concerned, direct the Data Subject to you.
9.3 Assessments and consultation. We will provide reasonable assistance with your obligations under Articles 32 to 36 GDPR, again taking into account the nature of the Processing and the information available to us. That includes data protection impact assessments — Annexes I(B), II and VI are written to be usable directly in a DPIA — prior consultation with a Supervisory Authority under Article 36, and securing the Processing and handling Security Incidents.
9.4 Cost. Assistance is included in your subscription. If requests are repetitive, manifestly excessive, or need substantial engineering work beyond the features of the Service, we may charge a reasonable fee at our standard hourly rate. We will tell you the likely cost and get your agreement before doing chargeable work.
10. Security incidents
10.1 We will notify you of a Security Incident affecting Customer Personal Data without undue delay and in any event within 48 hours of becoming aware of it. We “become aware” when we have a reasonable degree of certainty that such an incident has occurred; time spent establishing that certainty is not delay.
10.2 Our notice will, so far as known at the time, describe: (a) the nature of the incident, including where possible the categories and approximate number of Data Subjects and records concerned; (b) the likely consequences; (c) the measures taken or proposed to address it, including to mitigate adverse effects; and (d) a point of contact for more information. Where we cannot give all of that at once, we will provide it in phases as it becomes available, without further undue delay.
10.3 We notify the account owner and any administrator contact on your account, plus any additional security notice address you register with us. Please keep those addresses current — email privacy@bidwhistle.com to add or change one.
10.4 No admission. Our notice, and any step we take to investigate or remediate, is not an acknowledgement of fault or liability.
10.5 We will cooperate with you and give reasonable assistance with your obligations under Articles 33 and 34 GDPR, including any notification you must make to a Supervisory Authority or to Data Subjects.
10.6 We will not notify a Supervisory Authority or Data Subjects about a Security Incident affecting Customer Personal Data on your behalf unless you ask us to or the law requires it of us. Where we notify in our own right as a Controller, we will tell you.
10.7 We keep a record of Security Incidents affecting Customer Personal Data, including the facts, effects and remedial action, available to you on request under section 12.
10.8 If you find a security vulnerability, please tell us at security@bidwhistle.com. We will not take legal action against anyone who reports in good faith, gives us a reasonable chance to fix the problem, and does not access or alter other people’s data.
11. Return and deletion
11.1 At the end of the Agreement we will, at your choice, delete or return Customer Personal Data and delete existing copies. Tell us which before the end of the retrieval period; if you do not, we will delete.
11.2 Retrieval window. For 90 days after termination or expiry you can export Customer Data from your account in a structured, commonly used, machine-readable format. We do not charge for export, switching assistance or egress. This mirrors the switching and data portability section of the Terms of Service.
11.3 Deletion. We delete Customer Personal Data from live systems 90 days after termination or expiry — at the end of the retrieval window in clause 11.2 — whether the subscription was not renewed or you closed the account, or within 30 days of your written deletion instruction during the subscription, and we instruct our Sub-processors to do the same.
11.4 Backups. Encrypted backups are purged on their normal cycle and in any event within 30 days of deletion from live systems. Until then they are isolated, protected by the measures in Annex II, and used only for restoration.
11.5 Exceptions. We keep Customer Personal Data after termination only where required by Union or Member State law, or where it sits in the records below. Anything kept stays subject to this DPA and is Processed only for the purpose that justifies keeping it.
| Record | Retained for | Why |
|---|---|---|
| Invoices and accounting records | 7 years from the end of the financial year | Estonian Accounting Act § 12 |
| Security and access logs | 12 months | Security, fraud prevention, incident investigation |
| Support correspondence | 3 years | Follow-up queries and disputes |
| Data protection complaints and DSAR records | 3 years | Demonstrating compliance |
| Aggregated and de-identified insights | Indefinitely | Not personal data; subject to the covenant in section 4.5(c) |
| Anything under legal hold or subject to a claim | Until the hold or claim ends | Establishing, exercising or defending legal claims |
11.6 On written request we will confirm in writing that deletion has taken place.
12. Audits
12.1 We will make available the information reasonably necessary to demonstrate compliance with Article 28 GDPR and this DPA.
12.2 Documentation first. In the first instance we do that by providing this DPA (including Annexes II and VI), our current security documentation and the relevant parts of our Information Security Policy, our Sub-processors’ audit reports and certifications where they allow us to share them, and a completed security questionnaire. We will complete one reasonable questionnaire for you in any 12-month period, free of charge. We are a small company, so please use an industry-standard questionnaire where you can.
12.3 Audits. If that documentation does not reasonably answer your questions, you or an independent auditor you appoint may audit our compliance, subject to all of the following:
- not more than once in any 12-month period;
- on at least 30 days’ written notice to legal@bidwhistle.com;
- during our normal business hours, with reasonable steps to minimise disruption;
- limited to systems, documents and processes relevant to the Processing of your Customer Personal Data;
- subject to the auditor signing a reasonable confidentiality undertaking;
- not extending to another customer’s data, to commercially confidential information not relevant to data protection, or to physical inspection of a Sub-processor’s premises — for Sub-processors we provide their own audit reports and certifications instead; and
- at your cost, including our reasonable costs of supporting the audit.
12.4 You may audit more often, and paragraph 12.3(a) does not apply, where a Supervisory Authority requires it or following a Security Incident affecting your Customer Personal Data.
12.5 We may reasonably object to an auditor who is a competitor of ours or is not suitably qualified or independent, in which case you may appoint another.
12.6 If an audit reveals a material breach of this DPA by us, we will bear our own costs of that audit and remedy the breach without undue delay at our expense.
12.7 Audit findings are our confidential information and may be used only to assess our compliance with this DPA and to meet your obligations to a Supervisory Authority.
13. International transfers
13.1 Where the data actually goes, and which legs need a mechanism
13.1.1 We are established in Estonia, so for transfers out of the EEA we are the exporter.
13.1.2 Our system of record is in the United Kingdom. The platform that holds account data and Customer Data is hosted in London (Google Cloud europe-west2), so the main storage leg is an EU-to-UK transfer. The European Commission renewed its adequacy decisions for the United Kingdom on 19 December 2025, valid until 27 December 2031 with a review at the four-year point, so that leg is an adequacy transfer and needs no standard contractual clauses. Xano, Inc. is itself a US company: its data processing addendum includes the EU standard contractual clauses, and it is certified under the EU–US Data Privacy Framework and its UK Extension, which covers any access from the United States. Our supplier has not yet confirmed the London location in writing, including for backups. If the location proves to be elsewhere, we will put an appropriate mechanism in place and update Annex III and the Sub-processors page.
13.1.3 A UK customer sending data to us needs no mechanism. If you are established in the United Kingdom, transferring Customer Personal Data to us in Estonia is covered by the United Kingdom’s own adequacy regulations for the EEA. It is not a Restricted Transfer and neither the UK Addendum nor any other mechanism is required for it. The same is true for a customer established in Switzerland, whose transfer to us is covered by the Swiss adequacy recognition of the EEA.
13.1.4 The legs that do need a mechanism are the transfers to Sub-processors that process data in the United States: WeWeb, Anthropic, Voyage AI, Resend, Stripe and Twilio, and — through Anam — the Google model and Cartesia. That is where the SCCs, the UK Addendum, the Swiss overlay and any reliance on the Data Privacy Framework operate. Hetzner (Germany) is inside the EU and needs nothing, and Hostinger (Cyprus) uses the SCCs for anything it holds outside the EEA. Our transfer to Anam, a company established in the United Kingdom, is covered by the adequacy decisions in 13.1.2; Anam’s own onward transfers are covered in 13.7.
13.1.5 A transfer from you to us can still require a mechanism where you are established in a third country whose law does not recognise the EEA as adequate. Where that is so, this section applies to that leg too, with you as exporter and us as importer.
13.2 EU standard contractual clauses
13.2.1 Where Chapter V GDPR requires a transfer mechanism, the SCCs apply, in full and unamended except as completed below. They operate in two places, and it is worth being clear which is which:
- Between us and each Sub-processor, for the onward transfers in 13.1.4. We enter into them with the Sub-processor before Customer Personal Data reaches it, as your Processor. You are not a party to those clauses; we remain liable to you for the Sub-processor’s performance under section 8.8, and we will provide the agreed data protection terms on request.
- Between you and us, for the residual leg in 13.1.5. There the SCCs are deemed incorporated into this DPA and entered into by both of us, with no separate signature needed.
13.2.2 The module follows the leg of the transfer, and applies automatically with no further action needed:
| Leg | Exporter | Importer | Module |
|---|---|---|---|
| From us to a Sub-processor outside the EEA — the legs identified in 13.1.4, and in practice the only ones that arise | Us, acting as your Processor (or Sub-processor, where you are a Processor) | The Sub-processor | Module Three (processor to processor) |
| From you to us, in the rare case described in 13.1.5 where you are a Controller | You | Us | Module Two (controller to processor) |
| From you to us, in the rare case described in 13.1.5 where you are a Processor | You | Us | Module Three (processor to processor) |
For the avoidance of doubt, the onward transfers in 13.1.4 are made by us as a processor, so Module Three governs them whether you are a Controller or a Processor.
13.2.3 The SCCs are completed as follows:
| SCC provision | How it is completed |
|---|---|
| Clause 7 (docking clause) | Included |
| Clause 9(a) (sub-processors) | Option 2 — general written authorisation, notice period 30 days, applied as in section 8 |
| Clause 11(a) (redress) | The optional independent dispute-resolution wording is not included |
| Clause 13 and Annex I(C) | Competent supervisory authority: the Estonian Data Protection Inspectorate — Annex I(C) |
| Clause 17 (governing law) | Option 1 — the law of Estonia |
| Clause 18(b) (choice of forum) | The courts of Estonia |
| Annex I(A), I(B), II and III | Annexes I(A), I(B), II and III of this DPA |
13.2.4 If the SCCs are amended, replaced or invalidated, we will implement the replacement or other lawful mechanism, and this section is read accordingly.
13.3 United Kingdom
13.3.1 For a Restricted Transfer subject to the UK GDPR, the UK Addendum applies. It is deemed incorporated into this DPA, completed as set out in Annex IV, and it amends the SCCs as its Mandatory Clauses provide. In practice it covers the onward transfers in 13.1.4, where data originating in the United Kingdom, or subject to the UK GDPR, goes from us to a Sub-processor in a country not covered by UK adequacy regulations. It is not needed for a transfer from you in the United Kingdom to us in Estonia (13.1.3), and it is not needed for the storage leg to our UK-hosted platform, which is not a transfer out of the United Kingdom at all.
13.3.2 In Table 4 of the UK Addendum, “Importer” is selected as the party that may end the Addendum when the Approved Addendum changes.
13.4 Switzerland
13.4.1 For a Restricted Transfer subject to the FADP, the SCCs apply with the interpretive overlay in Annex V, Part 2 — including that the competent authority is the Swiss Federal Data Protection and Information Commissioner (FDPIC), that references to the GDPR are read as references to the FADP, and that Data Subjects in Switzerland may enforce their rights in Switzerland.
13.5 Data Privacy Framework
13.5.1 Where a Sub-processor in the United States is certified under the DPF for the relevant category of data, we may rely on that certification as the transfer mechanism in preference to the SCCs.
13.5.2 In every such case we also put the SCCs — and, for UK transfers, the UK Addendum — in place as a fallback, so that if the relevant adequacy decision is suspended, repealed or invalidated, or the Sub-processor’s certification lapses, the SCCs apply automatically and without interruption from that date.
13.5.3 Annex III and the Sub-processors page record which mechanism we currently rely on for each Sub-processor.
13.6 Transfer impact assessment and supplementary measures
13.6.1 We assess each Restricted Transfer against the laws and practices of the destination country, the nature of the data, the safeguards in place and the risk of government access. That assessment is recorded in 13.6.2 and 13.6.3 below and in Annexes III and VI. We have not yet produced it as a separate dated document; section 12 applies when you ask to see our records.
13.6.2 The supplementary measures we apply are: encryption in transit for all data leaving our environment; encryption at rest by the receiving platform; data minimisation, so AI Sub-processors receive only the content needed to answer the request and no account identifiers beyond what the provider requires; zero-retention or minimum-retention settings where the provider makes them available to us on our plan; contractual commitments from the importer to challenge unlawful access requests and notify us where permitted; and a preference for EU-hosted providers and EU data residency options when choosing new providers.
13.6.3 Two limits on 13.6.2 are worth stating rather than burying. First, our voice provider sells regional processing and zero data retention as account entitlements that are not available on the plan we hold, so for Olivia sessions we can apply neither: the region is the provider’s choice and may be the United States, and the session recording and transcript are held for 30 days (section 5.3.2 and Annex VI). Second, for the semantic retrieval provider the processing location is not published, and for the model reached through our voice provider we hold no location, retention or training commitment, so our assessment of those transfers rests on our own minimisation measures rather than on confirmed provider commitments. Third, our e-mail provider (Resend) and the host of our application front end (WeWeb) store data in the United States, so both transfers rely on the SCCs in their data processing terms and are covered by this assessment. The first two points are flagged in Annex VI and on the Sub-processors page, and both are on our pre-launch action list.
13.6.4 Where an assessment shows a transfer cannot be made with an adequate level of protection, we will not make it, and we will look for an alternative provider or region.
13.7 Onward transfers by sub-processors
13.7.1 We require each Sub-processor to make onward transfers only with a valid transfer mechanism in place and on terms no less protective than this DPA.
13.7.2 One onward transfer in our chain is significant enough to name. The reasoning model behind Olivia is a Google model that we do not contract for directly: our voice provider, Anam, contracts for it, so the terms covering that model are Anam’s rather than ours, and they are Google’s standard terms. We hold no zero-retention or no-training commitment covering it. The same is true of Cartesia, which generates Olivia’s voice for Anam: its data processing addendum uses the SCCs and the UK Addendum and also refers to the Data Privacy Framework, but Cartesia does not appear on the Framework’s official list (checked 28 September 2026). Anam’s own data processing addendum relies on the United Kingdom’s international data transfer tools for its onward transfers. We say so here, in Annex VI and on the Sub-processors page rather than leaving you to discover it. We hold no written confirmation of the transfer mechanism, retention or training position for the model reached through the voice provider, and are seeking it.
13.8 Government access requests
13.8.1 If we receive a legally binding request from a public authority for Customer Personal Data, we will: (a) tell you as soon as possible unless prohibited; (b) where prohibited, use reasonable efforts to obtain a waiver and to tell you as much as we can, as soon as we can; (c) challenge the request where there are reasonable grounds to consider it unlawful, including by seeking interim measures; and (d) disclose only the minimum permissible on a reasonable interpretation of the request.
13.8.2 We document each request and make that documentation available to you and to the competent Supervisory Authority on request. We have received no such request to date.
14. Records of processing
14.1 We maintain a record of the categories of Processing carried out on your behalf under Article 30(2) GDPR. Annex I(B) reflects that record for the Service, and we will make it available to you or to a Supervisory Authority on request.
14.2 We will give reasonable assistance with your own record under Article 30(1), including by keeping Annexes I(B), II, III and VI accurate and current.
14.3 We are established in the Union, so no Article 27 GDPR representative is required. We offer the Service to people in the United Kingdom without a UK establishment, so Article 27 of the UK GDPR requires us to appoint a representative in the United Kingdom. Our representative is James Rehman, contactable at privacy@bidwhistle.com, appointed 25 October 2026 and published in our Privacy Notice, section 18.1.2.
14.4 We are not required to appoint a Data Protection Officer and have not appointed one. Data protection questions go to privacy@bidwhistle.com, monitored by the person responsible for data protection at Bidwhistle.
15. Liability
15.1 Each party’s liability under or in connection with this DPA is subject to the exclusions and the limit of liability in the Terms of Service.
15.2 The Terms of Service and this DPA are a single agreement for that purpose. There is one aggregate cap across both, not a separate cap for each — a claim under this DPA counts towards the same cap as a claim under the Terms of Service.
15.3 Nothing here limits or excludes liability that cannot lawfully be limited or excluded: death or personal injury caused by negligence; fraud or fraudulent misrepresentation; intentional breach or gross negligence; a party’s indemnity obligations; your obligation to pay fees; and any statutory rights of consumers.
15.4 Data subjects come first. Nothing in this DPA or the Terms of Service limits either party’s liability to a Data Subject, including under Clause 12 of the SCCs, or to a Supervisory Authority. Any limitation applies only as between you and us.
15.5 Where both of us are responsible for damage caused by Processing, each is liable for the part attributable to its own breach, and we will cooperate in good faith on apportionment. A party that has paid full compensation to a Data Subject may claim back from the other the part corresponding to that other party’s responsibility.
15.6 This DPA creates no indemnity beyond those in the Terms of Service.
16. General
16.1 Term. This DPA takes effect on the effective date of the Agreement and continues until we have deleted or returned all Customer Personal Data under section 11. Sections 6, 11, 12, 13, 15 and 16 survive the end of the Agreement, together with any provision that by its nature should survive.
16.2 Changes. We may update this DPA to reflect a change in Data Protection Law, guidance from a Supervisory Authority, a new or replacement transfer mechanism, a change of Sub-processors, or a change to the Service. No change will materially reduce the protection this DPA gives to Customer Personal Data. For any material change we give at least 30 days’ notice by email to the account owner and update the version and dates at the top of this document; Sub-processor changes follow section 8 instead. If a material change is one you cannot reasonably accept, tell us within the notice period; if we cannot resolve it, you may terminate the affected part of the Service and receive a pro-rata refund for the unused remainder of your Subscription Period.
16.3 Governing law and jurisdiction. This DPA is governed by the laws of Estonia, and the courts of Estonia have exclusive jurisdiction, with the Harju County Court as court of first instance — mirroring the Terms of Service. Where the customer is a consumer, nothing here deprives them of the protection of the mandatory provisions of the law of their country of residence, and they may bring proceedings in their local courts. Clauses 17 and 18 of the SCCs govern the SCCs themselves and prevail for matters arising under them; section 16 of the UK Addendum governs the UK Addendum.
16.4 Notices. Notices under this DPA are given by email:
| Purpose | Address |
|---|---|
| Data protection questions, sub-processor objections, DSAR assistance | privacy@bidwhistle.com |
| Signed copies, audit notices, formal legal notices | legal@bidwhistle.com |
| Security vulnerabilities and incident correspondence | security@bidwhistle.com |
We give notice to the account owner’s email address and any administrator or additional notice address on your account. Keeping those current is your responsibility.
16.5 If any provision is held invalid or unenforceable it is severed and the rest continues. A failure to enforce is not a waiver. No one other than you and us may enforce this DPA, except that Data Subjects may enforce the third-party beneficiary rights given to them by the SCCs and the UK Addendum. This DPA and its Annexes are the entire agreement between us on the Processing of Customer Personal Data.
Annex I(A) — List of parties
This Annex is Annex I.A for the SCCs and supplies the “Parties” information for Table 1 of the UK Addendum.
Read it with section 13.1. Part 2 does the work: the transfers that actually need a mechanism are the onward ones from us to a Sub-processor outside the EEA and the United Kingdom. Part 1 is completed for the residual case in section 13.1.5, where you are established in a third country whose law does not treat the EEA as adequate; a customer in the United Kingdom or Switzerland does not need it, because their own jurisdiction’s adequacy findings for the EEA cover the transfer to us.
Part 1 — Transfers from you to us (residual case only — section 13.1.5)
Complete the exporter box when you sign a copy of this DPA. If you do not sign a copy, the details in your account record apply.
Data exporter — the Customer
| Item | Detail |
|---|---|
| Name | As in your account record, or as completed on a signed copy |
| Registered address | As in your account record, or as completed on a signed copy |
| Company or registry number | As in your account record, or as completed on a signed copy |
| Contact person — name, position, email | The account owner, or the data protection contact you give us |
| Activities relevant to the transferred data | Use of the Bidwhistle Service to search, assess, track and manage public procurement opportunities, including AI Features, Olivia and the Response Workspace |
| Role | Controller (Module Two) or Processor (Module Three), as set out in section 3 |
| Signature and date | Not required: this DPA takes effect on acceptance of the Terms of Service (section 2.1). A signed copy is available under section 2.2. |
Data importer — Bidwhistle OÜ
| Item | Detail |
|---|---|
| Name | Bidwhistle OÜ |
| Registry code | 17567745 (Estonian Commercial Register, Tartu County Court registration department) |
| Registered address | Sepapaja tn 6, 15551 Tallinn, Estonia |
| Contact person — position and email | Person responsible for data protection — privacy@bidwhistle.com |
| Activities relevant to the transferred data | Providing the Bidwhistle procurement intelligence platform: hosting the application, storing Customer Data, running searches, matching, alerts and fit scoring, operating Olivia (our real-time AI assistant with a synthetic voice and animated avatar), operating the Response Workspace, in which a customer uploads a tender pack and can have it read, have answers drafted, have a requirement explained and have a draft bid reviewed, and running semantic search over our knowledge base |
| Role | Processor (Module Two) or Sub-processor (Module Three), as set out in section 3 |
| UK Article 27 representative | James Rehman, privacy@bidwhistle.com — appointed 25 October 2026 and published in our Privacy Notice, section 18.1.2 |
| Signature and date | Entered into on acceptance of the Terms of Service; countersigned copy available from legal@bidwhistle.com. Signatory: James Rehman, sole member of the management board |
Part 2 — Transfers from us to a sub-processor (the transfers that need a mechanism)
Where we transfer Customer Personal Data to a Sub-processor outside the EEA and the United Kingdom, we are the data exporter (details as in the importer box above) and the Sub-processor is the data importer. Module Three applies, because we export in our capacity as your Processor.
The importers are the providers identified in section 13.1.4: Anam, Google (reached through Anam, as an onward transfer made under Anam’s own contract — see section 13.7.2), Anthropic, Voyage AI and Stripe. Each acts as a processor. Each one’s purpose, the data it receives, its processing location, its retention and training position and the transfer mechanism relied on are set out in Annex III, with the AI providers described in more detail in Annex VI, and the same information is published on the Sub-processors page.
Annex I(B) — Description of the transfer
1. Categories of data subjects
- Your Authorised Users — the people at your organisation who hold a Bidwhistle account or use the Service under your account
- Your own personnel and contacts whose details you record in Customer Data, for example in notes or bid team assignments
- Individuals mentioned by an Authorised User in an Input, including in a question put to Olivia
- Your customers’, clients’ or partners’ personnel, where you choose to put their details into the Service
- Third parties named in a document you upload to the Response Workspace — people who have no relationship with you or with us, but whose personal data appears in a tender pack. In practice this most often means the incumbent supplier’s employees named in a TUPE schedule, with roles, salaries and length of service, and named individuals in buyer correspondence, references, CVs and organisation charts included in the pack. You are the Controller of that data (section 3.7) and it reaches us because you upload it
Named public-sector contacts in Procurement Data are not in this list. We process them as an independent Controller under section 3.5, not on your behalf.
2. Categories of personal data
| Category | Examples |
|---|---|
| Identity and contact data | Name, job title, employer, work email address, work telephone number |
| Account data | User ID, role and permissions, preferences and settings |
| Authentication data | Hashed passwords or federated identity tokens, multi-factor authentication state, session data |
| Content you create | Saved searches and filters, alert rules, notes, tags, tracked tenders, pipeline entries, bid answers and draft responses, price lines, requirement lists and bid assessments |
| Uploaded tender pack contents | Whatever is in the document you upload to the Response Workspace: the tender documents themselves, and any personal data they carry — most significantly TUPE and staff schedules naming the incumbent supplier’s employees with roles, salaries and length of service, and named individuals in correspondence, references, CVs and organisation charts. We do not choose what is in the pack; you do |
| AI interaction data | Inputs, Outputs, questions put to Olivia, Olivia conversation transcripts, and the session recording (your Authorised User’s voice audio and the avatar video) |
| Communications | Support tickets and correspondence with us |
| Technical data | IP address, device and browser information, timestamps, log and event data linked to an account |
3. Special categories of data
None are intended, and you must not submit them. The Service is not designed for the special categories of personal data in Article 9 GDPR, or for data relating to criminal convictions and offences under Article 10. You must not submit such data, including through an Input, a question to Olivia, or a document you upload to the Response Workspace. If it is submitted despite this, we will process it under the same measures as all other Customer Personal Data — we do not apply additional safeguards designed for special category data — and you remain responsible for the consequences.
Uploads deserve a specific word. A tender pack is assembled by someone else and you may not have read every file in it. Section 4.8 makes checking it your responsibility, and section 3.7(d) records the controls we apply to help you do so — and their limit, which is that they read file names rather than file contents.
4. Frequency of the transfer
Continuous, for the duration of the Agreement.
5. Nature and purpose of the processing
We Process Customer Personal Data to provide the Service: creating and administering accounts and authenticating Authorised Users; storing and retrieving Customer Data; running searches, saved searches and matching, and sending email, text-message and in-app alerts; scoring tenders against your profile in our own ingestion and scoring pipeline, using fixed rules and no third-party model; operating Olivia, which means converting speech to text, injecting the profile context needed to answer, generating a response, converting text to synthetic speech and rendering an animated avatar in real time; operating the Response Workspace, which means passing an uploaded tender pack and your instructions to the AI Sub-processor identified in Annex VI so that the pack can be read, an answer drafted, a requirement explained or a draft bid reviewed, and returning the result to you; running semantic search over our knowledge base, which means turning your question into a vector at the provider identified in Annex VI; providing support you ask for; securing the Service and preventing and investigating abuse and fraud; and backing up and restoring data. Storage, structuring, retrieval, consultation, use, disclosure to Sub-processors, erasure and destruction are all included.
6. Retention
| Data | Retention |
|---|---|
| Account and profile data | For the life of the account. Deleted 90 days after the account ends, whether the subscription was not renewed or you closed the account |
| Customer Data (saved searches, notes, tracked tenders, bid answers, draft responses) | For the life of the account. Deleted 90 days after the account ends, whether the subscription was not renewed or you closed the account |
| Data deleted on request during the subscription | Removed from live systems within 30 days |
| Encrypted backups | Purged within 30 days of deletion from live systems |
| Documents you upload to the Response Workspace | Not stored by us. The pack is passed through to the AI Sub-processor to answer your request and discarded; only a usage row — token counts and a reference, no content — is written, and that row is kept as a product and usage log |
| Olivia session recordings (your Authorised User’s voice audio and the avatar video) and transcripts | Held by our voice provider for 30 days from the session, then deleted automatically; we also delete each session at the provider ourselves once it is 30 days old. Neither the provider’s zero-retention option nor its region selection is available on the plan we hold (section 5.3.2) |
| Olivia transcripts held in the platform | We have not established whether the platform keeps its own copy of Olivia transcripts and conversation logs, or for how long. We are establishing it and will record it here. |
| Platform request history — request and response bodies under roughly 10 KB, recorded by our application backend | Approximately 1 day (observed as about 2,200 requests at current traffic), then overwritten. Item 9 of Annex II explains what this covers and what we are doing about it |
| Voice embeddings / derived voice data | Not created or stored by us |
| Product and usage logs | 12 months |
| Security and access logs | 12 months |
| Free trial accounts that never convert | Deleted 90 days after the trial ends |
| Support correspondence | 3 years |
| Data protection complaints and DSAR records | 3 years |
| Invoices and accounting records | 7 years from the end of the financial year (Estonian Accounting Act § 12) |
7. Transfers to sub-processors
Each Sub-processor Processes the categories of Customer Personal Data described for it in Annex III and Annex VI, for the purpose stated there, for as long as we use it to provide the Service and for the retention period stated. AI and inference Sub-processors receive only the content needed to answer the request in hand and process it to return a result — except that a document you upload to the Response Workspace is sent in full, because answering a question about a tender pack requires the pack.
Annex I(C) — Competent supervisory authority
| Item | Detail |
|---|---|
| Authority | Estonian Data Protection Inspectorate (Andmekaitse Inspektsioon) |
| Address | Tatari 39, 10134 Tallinn, Estonia |
| Telephone | +372 627 4135 |
| info@aki.ee | |
| Website | www.aki.ee |
Where we are the data exporter for a transfer to a Sub-processor outside the EEA, the Estonian Data Protection Inspectorate is the competent supervisory authority, because we are established in Estonia.
Where you are the data exporter and are established in an EEA state, the competent authority is the authority of that state. Where you are not established in the EEA but the GDPR applies to you under Article 3(2), it is the authority of the Member State in which your Article 27 representative is established.
Individuals in the United Kingdom may also complain to the Information Commissioner’s Office (ICO), Wycliffe House, Water Lane, Wilmslow, Cheshire SK9 5AF, telephone 0303 123 1113, www.ico.org.uk. Individuals in Switzerland may complain to the Federal Data Protection and Information Commissioner (FDPIC).
Annex II — Technical and organisational measures
This Annex is Annex II for the SCCs and describes our measures under Article 32 GDPR.
Please read the honest version. Bidwhistle is a small, cloud-native company. We operate no data centre, we run almost no infrastructure of our own — one hosted server for the ingestion pipeline is the exception — we have no security team, and we hold no security certification. Our security model is to use reputable managed platforms, configure them carefully, keep the number of people and systems with access small, and be candid about the gaps. Anything not yet in place is marked planned, not yet in place. Item 9 records a logging behaviour that works against us; we would rather you read it here than find it later.
1. Pseudonymisation and encryption of personal data - All data in transit is encrypted with TLS 1.2 or above; HTTPS is enforced and plain HTTP is redirected. - All data at rest is encrypted by the underlying managed platforms — the application backend that is our system of record, its backups, and the ingestion host — using the encryption those platforms provide. - Passwords are never stored in plain text; the authentication service built into our application backend stores them as salted hashes. - Secrets and API keys are held in managed secret stores or a password manager with multi-factor authentication, never in source code or shared documents. - Requests to AI Sub-processors carry only the content needed to answer them, without billing data or contact details, and without account identifiers unless the provider requires one. The exception is a document you upload to the Response Workspace, which is sent in full because the request cannot be answered otherwise. - Systematic pseudonymisation of identifiers in log pipelines: planned, not yet in place.
2. Confidentiality, integrity, availability and resilience of processing systems - Access to production data is limited to the few named individuals who need it; every administrative account requires multi-factor authentication and there are no shared logins. - Changes are made through the platforms’ own change and publishing workflows, with development separated from production. - We rely on managed platforms that run their own redundant infrastructure rather than on data centres we operate. One component, the tender ingestion and scoring pipeline, runs on hosted infrastructure in Germany rather than on a fully managed platform; it holds no customer data at rest beyond logs. Availability commitments are in the Service Level Agreement. - AI Features degrade gracefully: if a model provider is unavailable, search, saved searches, alerts and tracked tenders keep working. - Formal capacity and resilience testing: planned, not yet in place. - A documented patching and hardening routine for the self-administered ingestion host: planned, not yet in place.
3. Ability to restore availability and access in a timely manner - Customer Data is held in managed platforms that take regular automated backups, encrypted at rest by the platform, from which an environment can be restored. - Documented recovery point and recovery time objectives, and a scheduled restore test at least annually: planned, not yet in place. - A written business continuity and disaster recovery plan: planned, not yet in place.
4. Regular testing, assessing and evaluating effectiveness - We review access rights, active integrations and the Sub-processor list at least every six months. - We keep dependencies and platform configurations current, apply providers’ security updates, and run automated vulnerability alerts on repositories we control. - We operate a vulnerability disclosure route at security@bidwhistle.com and investigate reports in good faith. - We have not carried out an independent penetration test and do not claim to have done so. An external security review: planned, not yet in place. - A formal annual review of this Annex against the risks we face: planned, not yet in place.
5. User identification and authorisation - Each Authorised User has an individual account; credential sharing is prohibited by the Terms of Service and the Acceptable Use Policy. Role-based permissions govern what each user can see and do. - Administrative and platform accounts require multi-factor authentication, follow least privilege, are reviewed at least every six months, and are revoked the same day an engagement ends. - Session expiry and re-authentication are enforced by the authentication built into our application backend. - Single sign-on and SCIM provisioning for customers: planned, not yet in place.
6. Protection of data during transmission - TLS 1.2 or above for all connections between browsers, our application, our backend and our Sub-processors, with HTTP Strict Transport Security enabled. - Olivia’s real-time audio and video use the encrypted transport of the voice and avatar provider named in Annex III. - Keys and tokens travel only over encrypted channels and are rotated when someone with access leaves or we suspect exposure. - We do not send Customer Personal Data by unencrypted email, and ask you not to send it to us that way.
7. Protection of data during storage - Customer Data is stored in the single managed application backend named in Annex III, which is our system of record, encrypted at rest by that platform. We do not keep a second copy of Customer Data in another store. - Documents you upload to the Response Workspace are not stored: they are passed through to the AI Sub-processor to answer your request and discarded, and only a usage row with no content is written. - The front end and the Olivia embed are static delivery: your data passes between your browser and our system of record rather than resting in either, and your session token lives in your own browser. The front end is hosted by WeWeb in the United States (Annex III). - Production and non-production environments are separated; we do not use real Customer Personal Data in test or demonstration environments. - Backups are encrypted and purged within 30 days of deletion from live systems. Retention limits are applied as in Annex I(B). Customer-managed encryption keys are not offered. - Our platform supplier has not yet confirmed the hosting region in writing, including for backups.
8. Physical security of processing locations - We operate no data centre or physical infrastructure. All Processing happens in facilities run by our Sub-processors, which are responsible for physical security and publish their own measures. - Our working environment is remote. Devices used for production access have full-disk encryption, automatic screen lock, current operating systems and endpoint protection. No paper records are created. - A formal written device and asset register: planned, not yet in place.
9. Event logging — including one thing we would rather you heard from us - The application backend records authentication events, administrative actions and errors. Access to logs is limited to administrators. - Security and access logs are kept 12 months; product and usage logs 12 months. - Request history. Our application backend has a request-history feature that records the body sent to an endpoint and the body returned, verbatim, while the serialised value stays under roughly 10 KB; above that threshold the whole object is discarded rather than truncated. Observed retention is about 2,200 requests, which at current traffic is roughly one day, after which the oldest entries are overwritten. It is on by default. We have switched it off on 68 of our 178 endpoints; the remaining 110 inherit it, and those include endpoints that carry a customer’s actual bid text — saving an answer, saving the response, saving requirements, saving price lines, saving the bid assessment, and the three that call the model. Reading an uploaded pack is not affected: an uploaded document is far above the 10 KB threshold, so its content was never retained by this feature. - We record this because it is a real retention behaviour affecting Customer Personal Data, it is not obvious from the outside, and a buyer assessing us is entitled to know it. It also appears in the retention table in Annex I(B), in our Data Retention Policy and in our Privacy Notice. - Planned, not yet done: request-history logging to be disabled on the endpoints carrying customer bid text - Centralised log aggregation with alerting on anomalous administrative activity, and tamper-evident log storage: planned, not yet in place.
10. System configuration, including default configuration - We prefer secure managed-platform defaults and change them only with a reason. Data stores are private by default; public sharing links and open endpoints are not used for Customer Data. - Provider settings that affect data protection — data residency, retention, and AI training or feedback settings — are set deliberately and recorded in Annex VI. Where a provider does not offer the setting to us on our current plan, Annex VI says so rather than leaving the impression that we chose the weaker option: that is the position today for regional processing and zero data retention with our voice provider (section 5.3.2). - New integrations are added only where needed and with the minimum scopes required. - A written configuration baseline for each platform: planned, not yet in place.
11. Internal IT and IT security governance - One named person is accountable for information security and data protection. We do not call anyone a Data Protection Officer, because we are not required to appoint one. - Our internal Information Security Policy and Data Retention Policy set out how we work; they are written for a small team and honest about it. - Security incidents follow a documented process: detect, contain, assess, notify, remediate, record, review. - Before a new feature or Sub-processor is added we check the provider’s terms, security documentation, retention behaviour and AI training position. Contractors are engaged under written confidentiality and data protection terms. - A formal risk register and scheduled management review: planned, not yet in place.
12. Certification and assurance - We hold no security certification. We are not certified to ISO/IEC 27001, SOC 2, Cyber Essentials or any equivalent scheme, and we will not claim otherwise. - We rely instead on our Sub-processors’ certifications and audit reports, which we review when selecting a provider and when reviewing the Sub-processor list, and share where the provider allows. - Pursuing a formal certification: planned, not yet in place.
13. Data minimisation - We collect only what we need to run the Service; account creation asks for the minimum needed to create and bill an account. - AI Sub-processors receive only the content needed to answer the request in hand — except a document you upload to the Response Workspace, which is sent in full because the request cannot be answered otherwise (Annex VI, paragraph 2.2). - We do not require personal telephone numbers or home addresses from Authorised Users. - We do not create voice embeddings or other derived biometric data, and Olivia’s voice is not a clone of any identifiable person’s voice. - We do not store documents uploaded to the Response Workspace; they are passed through and discarded. - We use no separate third-party product analytics provider at present. Any non-essential cookie or analytics use would run only with consent, as described in our Cookie Notice. - The one place where minimisation is currently weaker than we want is the request-history behaviour in item 9, and the remedy is flagged there.
14. Data quality - Account and profile data can be corrected at any time from within the Service. - Customer Data is what you put in; we do not alter it except as you instruct through the features of the Service. - Outputs are generated by AI and can be wrong. We say so in the product and in our AI Use Disclosure, and we tell users to verify material facts against the original notice. - Procurement Data is reproduced from official sources and only the notice published by the contracting authority is authoritative. Personal data within it is reviewed annually, and published contact details are removed 24 months after the relevant tender’s closing date unless still needed.
15. Limited data retention - Retention periods are in Annex I(B) and our Data Retention Policy, applied consistently with the Privacy Notice. - Deletion on request completes in live systems within 30 days, with backups purged within 30 days after that. Free trial accounts that never convert are deleted 90 days after the trial ends. - Automated enforcement of every retention period, rather than a mix of automated and manual: planned, not yet in place.
16. Accountability - We maintain an Article 30(2) record of processing, a Legitimate Interests Assessment for Procurement Data, and a transfer impact assessment for Restricted Transfers. - We keep and publish a Sub-processor register and notify changes. - We keep records of Security Incidents, data subject requests and complaints, and answer one reasonable security questionnaire per customer per year.
17. Data portability and erasure - You can export Customer Data at any time, and for 90 days after termination, in a structured, commonly used, machine-readable format. We charge nothing for export, switching assistance or egress. - Deletion is available in-account and by writing to privacy@bidwhistle.com, within the periods in section 11, and we instruct Sub-processors to delete on the same basis.
18. Measures applied to transfers to sub-processors - We select Sub-processors offering appropriate security and, where possible, EU hosting or EU data residency. - Each is engaged under a written contract with data protection terms no less protective than this DPA. For AI Sub-processors we seek a training prohibition; section 5.2 records which providers have given us one and which have not, and we do not claim a prohibition we do not hold. - Transfers outside the EEA and the United Kingdom use the SCCs, the UK Addendum, or the DPF with SCCs as fallback, as recorded in Annex III. The main storage leg is an EU-to-UK adequacy transfer and needs no mechanism (section 13.1.2). - We apply the supplementary measures in section 13.6 and review Sub-processors’ security documentation at least annually and whenever we add one.
Covenant on changes. We may update these measures from time to time, provided the updates do not materially reduce the overall level of security of the Service.
Annex III — Sub-processors
The authoritative and current list is published at Sub-processors, where you can subscribe to change notifications. Section 8 governs how we add, notify and handle objections to Sub-processors. The list below reproduces that page as at the effective date, and the two are kept identical.
The processing location column records where the data is processed, which is what a transfer assessment turns on. Where the entity we contract with sits in a different country, the contract is with the entity named in that provider’s own data processing terms. They are listed below, as each provider’s own terms named them on 28 September 2026; registration numbers are shown where the provider publishes them.
- Xano — Xano, Inc., 21600 Oxnard Street, Suite 910, Woodland Hills, California 91367, United States.
- Hetzner — Hetzner Online GmbH, Industriestr. 25, 91710 Gunzenhausen, Germany; Ansbach commercial register HRB 6089.
- WeWeb — WeWeb, Inc., 251 Little Falls Drive, Wilmington, Delaware 19808, United States, as named in its data processing agreement. WeWeb’s Terms and Conditions name WEWEB SAS, 128 rue La Boétie, 75008 Paris, RCS 839 313 848. We are settling which entity we contract with.
- Vercel — Vercel Inc., 440 N Barranca Ave #4133, Covina, California 91723, United States; Delaware registration 5857312.
- Anam — ANAM.AI LTD, White Collar Factory, Old Street Yard, London EC1Y 8AF, United Kingdom; company number 15214363.
- Google and Cartesia — contracted by Anam, not by us. Cartesia is Cartesia AI, Inc., 555 De Haro Street, Suite 300, San Francisco, California 94107, United States.
- Anthropic — Anthropic Ireland, Limited, 6th Floor, South Bank House, Barrow Street, Dublin 4, D04 TR29, Ireland.
- Voyage AI — Voyage AI Innovations, Inc., United States, part of MongoDB, Inc.; its address is not published.
- Resend — Plus Five Five, Inc., 2261 Market Street #5039, San Francisco, California 94114, United States.
- Stripe — Stripe Payments Europe, Limited, Ireland, joined by Stripe Technology Europe, Limited where Stripe’s service terms say so.
- Twilio — Twilio Ireland Limited, 70 Sir John Rogerson’s Quay, Dublin 2, D02 R296, Ireland.
- Hostinger — Hostinger International Limited, 61 Lordou Vironos str., 6023 Larnaca, Cyprus.
- Telegram and GitHub — handle no personal data.
A. Infrastructure and platform
| Sub-processor | Purpose | Data processed | Processing location | Transfer mechanism |
|---|---|---|---|---|
| Xano | System of record — application backend, authentication, and all stored Customer Data | Name, e-mail address, company, sector, preferences, saved tenders, bids, alerts, usage and login events | London, United Kingdom (Google Cloud europe-west2). Evidence-based; not confirmed in writing by the supplier, including for backups. | None required for the storage leg on the evidence we have: the European Commission renewed its adequacy decisions for the United Kingdom on 19 December 2025, valid until 27 December 2031, so an EU-to-UK transfer needs no further mechanism. Xano, Inc. is a US company: its data processing addendum includes the EU standard contractual clauses, and it is certified under the EU–US Data Privacy Framework and its UK Extension, which covers any access from the United States, including by its own support tools. If the location turns out to be elsewhere we will put a mechanism in place and update this page |
| Hetzner | Ingestion host — the tender ingestion and scoring pipeline | Reads customer profiles in order to score them against tenders; holds no Customer Data at rest beyond logs | Nuremberg, Germany (EU) | None required (EU). Hetzner’s data processing agreement keeps processing inside the EU and EEA. We are recording our acceptance of Hetzner’s data processing agreement in our account. |
| WeWeb | The application front end | Delivery of the interface. Your data passes between your browser and our system of record rather than resting with WeWeb, and your session token lives in your own browser | United States. WeWeb’s data processing agreement states that both its editor and the applications published with it are hosted on Amazon Web Services in the USA | EU standard contractual clauses, under WeWeb’s data processing agreement. WeWeb does not delete data automatically when a contract ends, so we will ask it to if we leave |
| Vercel | Hosting of the Olivia assistant embed | Static delivery of the embed; session context passes through the browser, not through storage | Vercel edge; region immaterial, nothing rests there | None required — no personal data at rest. Should that change, Vercel’s data processing addendum includes the EU standard contractual clauses and the UK IDTA, and Vercel Inc. is certified under the EU–US Data Privacy Framework |
B. AI and inference
| Sub-processor | Purpose | Data processed | Processing location | Retention | Uses data to train models? | Transfer mechanism |
|---|---|---|---|---|---|---|
| Anam | Olivia’s avatar, voice, speech-to-text and hosted reasoning | The Authorised User’s voice audio, the conversation transcript, and the profile context injected into the session | Anam runs production on CoreWeave in Spain (EU) and the eastern United States, and the region for a session is its choice. Region selection is an account entitlement not available on our current plan. Region selection and zero data retention are not available to us on the plan we hold. We are discussing both with the provider. | Session recording (audio and avatar video) and transcript kept 30 days, then deleted automatically; we also delete each session at Anam ourselves once it is 30 days old. Zero data retention is an account entitlement not available on our current plan | No — confirmed by the provider: conversations are not used for model training | None required for our transfer to Anam: ANAM.AI LTD is a UK company, and the European Commission’s adequacy decisions for the United Kingdom cover it. Anam’s own onward transfers, including to the United States, are made under the UK transfer tools named in its data processing addendum. |
| Google (reached through Anam) | The reasoning model behind Olivia, contracted through Anam and not by us directly | Whatever reaches the model during an Olivia conversation | Reached through Anam; processing location not confirmed to us | Runs under Google’s standard terms; no zero-retention commitment covering it | Not confirmed. Google’s standard terms give us no specific no-training commitment. The weakest data commitment in our chain (section 13.7.2) | Onward transfer made by Anam under its own contract. We hold no written confirmation of the transfer mechanism, retention or training position for the model reached through the voice provider, and are seeking it. |
| Cartesia (reached through Anam) | Olivia’s synthetic voice — turning her replies into speech — contracted through Anam rather than directly by us | The text of Olivia’s replies, as they are spoken | Not published. Cartesia AI, Inc. is a US company | Not stated to us | Not confirmed to us | An onward transfer made by Anam under its own contract. Cartesia’s data processing addendum uses the EU standard contractual clauses (Modules 2 and 3) and the UK Addendum. It also refers to the EU–US Data Privacy Framework, but Cartesia does not appear on the Framework’s official list (checked 28 September 2026) |
| Anthropic | The reasoning model behind the Response Workspace — reading an uploaded tender pack, drafting an answer, explaining a requirement, reviewing a draft bid | The uploaded document, the customer’s own bid text and their company name. A tender pack routinely carries personal data about other people | Stored in the United States, the only storage region Anthropic offers; processed in any region by default | Anthropic states that it deletes API inputs and outputs within 30 days, except where its terms allow longer — for example to enforce its usage policy or meet a legal requirement | No — Anthropic's Commercial Terms state: “Anthropic may not train models on Customer Content from Services.” | Our contract is with Anthropic Ireland, Limited. Its Data Processing Addendum, incorporating the EU Standard Contractual Clauses (Module 2 or 3) with the UK Addendum and a Swiss addendum, is accepted automatically together with its Commercial Terms. |
| Voyage AI | Semantic retrieval for the assistant’s knowledge base — turning a question into a vector | The question the Authorised User asks, typed or spoken. No account identity travels with it | Not published. Voyage AI is a US company, part of MongoDB, and names Amazon Web Services, Google and Baseten as its own providers, so we treat the processing as taking place in the United States. | Zero-day retention on its hosted model API endpoints — exercising the opt-out switches retention off as well as training. | No — Bidwhistle has exercised Voyage's opt-out (10 September 2026). ⚠️ Voyage's terms take a training licence BY DEFAULT unless the customer opts out, so this rests on our opt-out rather than on their default position. | Voyage AI’s data processing addendum includes the EU standard contractual clauses, the UK Addendum and a Swiss addendum, and MongoDB, Inc.’s EU–US Data Privacy Framework certification covers Voyage AI Innovations, Inc. Voyage AI was acquired by MongoDB on 24 February 2025, and the licence in its terms runs to its “successors and assigns”. |
We have not recorded “no training” against a provider that has not confirmed it. Section 5.2 explains the position and Annex VI sets out what each provider receives.
C. Business operations
| Sub-processor | Purpose | Data processed | Processing location | Transfer mechanism |
|---|---|---|---|---|
| Resend | Alert and transactional e-mail — tender alerts, renewal reminders, account and security notifications | E-mail address, name, and the tender content of the message | United States. Resend sends through Amazon SES in Ireland but stores message content, delivery logs, webhook payloads and account records in the United States, and keeps e-mail and log data for 30 days | EU standard contractual clauses (Module 2) with the UK Addendum, under Resend’s data processing addendum. Resend is also listed under the EU–US Data Privacy Framework (its re-certification was under review on 28 September 2026) |
| Stripe | Billing — card payments, subscriptions, invoicing and refunds | Name, e-mail address, billing address, payment method. Stripe holds the card details; we never do | United States (Stripe, LLC), under Stripe’s own terms | Stripe’s Data Transfers Addendum relies first on the EU–US Data Privacy Framework, under which Stripe, LLC is certified, and then on the EU standard contractual clauses and the UK Addendum. For fraud prevention and anti-money-laundering checks Stripe acts as a controller in its own right |
| Twilio | Text-message (SMS) alerts | Your mobile number and the text of the alert | Twilio’s default region is the eastern United States; its Dublin region also handles text messages. We have not confirmed which Twilio region our account uses. | Twilio’s data protection addendum relies on the EU–US Data Privacy Framework (Twilio Inc. is certified), then Twilio’s binding corporate rules, then the EU standard contractual clauses and the UK IDTA. Twilio also acts as a controller in its own right for some account and usage data |
| Hostinger | Our e-mail mailboxes — support, privacy, legal and security correspondence | Anything you write to us by e-mail, and our replies | Not published for a mailbox: Hostinger states data may be stored in the United Kingdom, the Netherlands, Lithuania or Cyprus, and in other places where necessary | EU standard contractual clauses (Module 2 or 3) with the UK Addendum, under Hostinger’s data processing agreement, for anything held outside the EEA |
| Telegram | Fault alerting — operational messages telling us something has broken | None. Operational messages only; listed for completeness | Global | None required — no personal data |
| GitHub | Source control | None, provided no Customer Data reaches the repository | Global | None required — no personal data |
Analytics. We use no separate third-party product analytics provider at present.
Support. Support is email only, through support@bidwhistle.com. Our mailboxes are hosted by Hostinger (Part C). We use no separate helpdesk product.
Sources, not sub-processors. The public procurement portals we collect Procurement Data from are sources, not Sub-processors: data flows one way, from them to us, and we send them nothing about you. They are therefore not listed here and no transfer mechanism arises for them.
Annex IV — UK International Data Transfer Addendum
This Annex completes the International Data Transfer Addendum to the EU Commission Standard Contractual Clauses, version B.1.0 (the UK Addendum), issued by the Information Commissioner under section 119A of the Data Protection Act 2018 and laid before Parliament on 2 February 2022.
What this Annex is for. It applies where a transfer of Customer Personal Data is a Restricted Transfer under the UK GDPR. In practice that is a single situation: we transfer Customer Personal Data that is subject to the UK GDPR onward to a Sub-processor in a country not covered by UK adequacy regulations — the providers identified in section 13.1.4 and listed in Annex III, Parts B and C. We are the exporter for those transfers and the Sub-processor is the importer.
What this Annex is not for. Two transfers that a reader might expect to find here do not need the Addendum at all:
- a transfer from you in the United Kingdom to us in Estonia is covered by the United Kingdom’s own adequacy regulations for the EEA (section 13.1.3); and
- the storage leg to our UK-hosted application backend is not a transfer out of the United Kingdom at all, and from the EU side it is covered by the European Commission’s renewed UK adequacy decisions of 19 December 2025 (section 13.1.2).
Neither needs a mechanism, and neither is completed below. The tables record how the UK Addendum is completed in the agreements we enter into with each Sub-processor, reproduced here so that you can see the terms on which your UK data travels onward. We agree to be bound by the Mandatory Clauses of the UK Addendum, which amend the SCCs so they work for UK transfers.
Table 1 — Parties
| Item | Exporter | Importer |
|---|---|---|
| Start date | The date the Sub-processor is engaged, and in any event before Customer Personal Data reaches it | The same date |
| Party’s legal name | Bidwhistle OÜ | The Sub-processor named in Annex III, Part B or Part C, for the transfer in question |
| Trading name (if different) | Bidwhistle | As stated in Annex III |
| Main address | Sepapaja tn 6, 15551 Tallinn, Estonia | The address in that Sub-processor’s own data processing terms, as listed in Annex III |
| Official registration number | 17567745 (Estonian Commercial Register) | As above |
| Key contact — name and position | Person responsible for data protection | The data protection contact given in that Sub-processor’s own data processing terms |
| Key contact — email | privacy@bidwhistle.com | As above |
| Signature | Entered into by us on engaging the Sub-processor; a copy of the data protection terms is available under section 8.8 | Entered into by the Sub-processor on the same terms |
Where you are a Controller or a Processor whose UK data we transfer onward, you are not a party to that Addendum: we make the transfer as your Processor, we remain liable to you for it under section 8.8, and we may provide the agreed data protection terms on request.
Table 2 — Selected SCCs, modules and selected clauses
| Item | Selection |
|---|---|
| Addendum EU SCCs | The Approved EU SCCs — the standard contractual clauses annexed to Commission Implementing Decision (EU) 2021/914 of 4 June 2021, including the Appendix Information, incorporated by section 13.2 and completed as below |
| Module in operation | Module Three (processor to processor) — we export as your Processor (or Sub-processor) and the Sub-processor imports as a processor. Module Two applies only in the residual case in section 13.1.5, where you are a Controller exporting to us |
| Clause 7 (docking clause) | Used |
| Clause 9(a) (authorisation of sub-processors) | Option 2 — general written authorisation. Notice period for sub-processor changes: 30 days, applied as in section 8 |
| Clause 11(a) (optional independent dispute resolution wording) | Not used |
| Clause 17 (governing law) | Option 1 — the law of Estonia |
| Clause 18(b) (choice of forum and jurisdiction) | The courts of Estonia |
| Is personal data received from the importer combined with personal data collected by the exporter? | No |
Table 3 — Appendix information
| Appendix information | Where it is set out |
|---|---|
| Annex 1A: List of parties | Annex I(A), read with Table 1 above |
| Annex 1B: Description of transfer | Annex I(B) |
| Annex II: Technical and organisational measures | Annex II |
| Annex III: List of sub-processors (Module Three) | Annex III, read with Sub-processors |
For UK transfers, references in the Appendix Information to the competent supervisory authority are read as references to the Information Commissioner.
Table 4 — Ending this Addendum when the Approved Addendum changes
| Which parties may end this Addendum as set out in Section 19 of the Mandatory Clauses | Selection |
|---|---|
| Importer | ☑ Selected |
| Exporter | ☐ Not selected |
| Neither party | ☐ Not selected |
On the transfers this Annex actually covers, the Importer is the Sub-processor and the Exporter is us (Table 1). Where the Addendum instead operates on the residual leg in section 13.1.5, the Importer is us.
Annex V — Jurisdiction-specific terms
Part 1 — United Kingdom
V.1.1 Where Customer Personal Data is subject to the UK GDPR, references in this DPA to the GDPR are read as references to the UK GDPR, read with the Data Protection Act 2018 as amended by the Data (Use and Access) Act 2025.
V.1.2 References to Union or Member State law are read as references to the law of the United Kingdom or a part of it, and references to a Supervisory Authority as references to the Information Commissioner.
V.1.3 Restricted Transfers under the UK GDPR are made under the UK Addendum, completed in Annex IV. In practice those are the onward transfers from us to a Sub-processor outside the United Kingdom’s adequacy regulations (section 13.1.4). A transfer from a customer in the United Kingdom to us in Estonia is not a Restricted Transfer, because the United Kingdom’s adequacy regulations cover the EEA (section 13.1.3), and our UK-hosted application backend is not a transfer out of the United Kingdom at all. Where we rely on the DPF for a US Sub-processor we rely on the UK Extension to the EU–US Data Privacy Framework, with the UK Addendum as the fallback under section 13.5.2.
V.1.4 Article 27 of the UK GDPR requires us to appoint a representative in the United Kingdom. Our representative is James Rehman, contactable at privacy@bidwhistle.com, appointed 25 October 2026 and published in our Privacy Notice, section 18.1.2.
V.1.5 Where we send direct marketing to UK recipients we comply with PECR as well as the UK GDPR.
Part 2 — Switzerland
V.2.1 Where Customer Personal Data is subject to the FADP, the SCCs apply with the following interpretation.
V.2.2 The competent supervisory authority is the Federal Data Protection and Information Commissioner (FDPIC) for transfers governed by the FADP alone. Where a transfer is governed by both the GDPR and the FADP, the authorities in Annex I(C) apply in parallel and the obligations are cumulative.
V.2.3 References to the GDPR are read as references to the FADP, and references to Union or Member State law as references to Swiss law.
V.2.4 References to a “Member State” must not be read so as to prevent Data Subjects habitually resident in Switzerland from bringing proceedings in Switzerland, in line with Clause 18(c) of the SCCs.
V.2.5 The revised FADP, in force since 1 September 2023, protects the personal data of natural persons. Where an obligation arose while the previous Act also protected data of legal entities, the SCCs are read as extending to that data for that period.
V.2.6 Where we rely on the DPF for a US Sub-processor in respect of data subject to the FADP, we rely on the Swiss–US Data Privacy Framework, with the SCCs as the fallback under section 13.5.2.
Part 3 — Other jurisdictions
V.3.1 We contract with customers worldwide from Estonia, and this DPA is drafted to the GDPR and UK GDPR standard, which in our experience meets or exceeds the processor obligations of most other regimes.
V.3.2 We do not currently offer jurisdiction-specific addenda for other regimes — for example the California Consumer Privacy Act, the Brazilian LGPD or the Australian Privacy Principles — and we do not claim compliance with them.
V.3.3 If your regulator or your own contracts require additional terms, email legal@bidwhistle.com. We will consider a reasonable request, but we will not agree terms we cannot actually keep.
Annex VI — AI processing description
This Annex describes how AI Features process Customer Personal Data. The SCCs do not require it. We publish it because a buyer assessing an AI-enabled service needs it, and because it is the part of this pack a reviewer should read first.
It is written to be checkable. Where a position is confirmed, it says so and by whom. Where it is not, it says that instead of filling the gap.
1. The two AI surfaces
1.1 Olivia — the AI concierge. An animated avatar with a synthetic voice. An Authorised User speaks or types; the audio is transcribed, the profile context needed to answer is injected, a model produces a reply, and the avatar speaks it. The avatar, the voice, the speech-to-text and the hosting of the reasoning are all provided by Anam; the reasoning model itself is a Google model that Anam contracts for, not us.
1.2 The Response Workspace — the bid-writing surface. An Authorised User uploads a tender pack and can have it read, have answers drafted, have a requirement explained and have a draft bid reviewed. This runs on Anthropic. The pack is passed through and discarded: we store none of it, and only a usage row — token counts and a reference, no content — is written.
1.3 Knowledge search. A question put to the assistant’s knowledge base is turned into a vector by Voyage AI so that the right material can be retrieved.
1.4 What reaches an AI provider is therefore not the same in each case, and the tables below say so provider by provider rather than in aggregate.
2. What each AI sub-processor receives
| Provider | What it receives | Purpose | Processing location | Retention position | Training position |
|---|---|---|---|---|---|
| Anam | The Authorised User’s voice audio, the conversation transcript, and the profile context injected into the session. Not your saved searches in bulk, not your billing data, not your customer list | Olivia’s avatar, voice, speech-to-text and hosted reasoning | Anam runs production on CoreWeave in Spain (EU) and the eastern United States, and the region for a session is its choice. Region selection is an account entitlement that is not available on the plan we hold, so we cannot pin it to the EU today. This is a capability the provider does not sell on our plan, not a setting we have failed to switch. Region selection and zero data retention are not available to us on the plan we hold. We are discussing both with the provider. | The session recording — the user’s audio and the avatar video — and the transcript are kept 30 days, then deleted automatically. Zero data retention is likewise an account entitlement not available on our plan | No. Anam has confirmed that it does not use conversations for model training |
| Google (reached through Anam) | Whatever reaches the model during an Olivia conversation: the transcribed question, the injected profile context and the conversation so far | The reasoning behind Olivia’s replies | Reached through Anam. The processing location is not confirmed to us | Runs under Google’s standard terms. We hold no zero-retention commitment covering it | Not confirmed. Google’s standard terms give us no specific no-training commitment. Of everything in this Annex, this is the weakest data commitment in our chain, and we would rather you heard it here |
| Cartesia (reached through Anam) | The text of Olivia’s replies, as they are spoken | Turning Olivia’s replies into her synthetic voice | Not published. Cartesia AI, Inc. is a US company | Not stated to us | Not confirmed to us |
| Anthropic | The uploaded tender pack in full, the Authorised User’s own bid text, the instruction given, and the customer’s company name. A tender pack routinely carries personal data about other people — most significantly a TUPE schedule naming the incumbent supplier’s staff | Reading an uploaded pack, drafting an answer, explaining a requirement, reviewing a draft bid | Stored in the United States, the only storage region Anthropic offers; processed in any region by default | Anthropic states that it deletes API inputs and outputs within 30 days, except where its terms allow longer — for example to enforce its usage policy or meet a legal requirement | No — Anthropic's Commercial Terms state: “Anthropic may not train models on Customer Content from Services.” |
| Voyage AI | The question the Authorised User asks, typed or spoken. No account identity travels with it | Turning a question into a vector for semantic retrieval over our knowledge base | Not published. Voyage AI is a US company, part of MongoDB, and names Amazon Web Services, Google and Baseten as its own providers, so we treat the processing as taking place in the United States. | Zero-day retention on its hosted model API endpoints — exercising the opt-out switches retention off as well as training. | No — Bidwhistle has exercised Voyage's opt-out (10 September 2026). ⚠️ Voyage's terms take a training licence BY DEFAULT unless the customer opts out, so this rests on our opt-out rather than on their default position. |
2.1 What is outstanding, in one place. Written confirmation of processing location, retention and training position is outstanding for the Google model reached through Anam, and Voyage AI does not publish where it processes data; and a plan that offers EU region selection and zero data retention is outstanding for Anam. We hold no written confirmation of training, retention or processing location for the Google model reached through Anam, and none of the processing location for Voyage AI. We are seeking both and will record them here. Anthropic’s position is taken from its published terms. Region selection and zero data retention are not available to us on the plan we hold. We are discussing both with the provider. We will update this Annex, Annex III and the Sub-processors page as each is answered, and notify subscribers under section 8.
2.2 The uploaded pack is the sensitive part. Everywhere else in the Service, an AI provider receives only what is needed to answer a request. The Response Workspace is different: answering a question about a tender pack requires the pack, so the whole document goes to Anthropic. You are the Controller of everything in it (section 3.7) and you warrant the upload is lawful (section 4.8). Only the files you leave ticked are transmitted, and files whose names indicate a staff record are unticked by default; that default is overridable and reads file names rather than contents, so what you finally tick is transmitted in full (section 3.7(d)).
2.3 We do not create voice embeddings, voiceprints or other derived biometric data. Olivia’s voice is synthetic and is not a clone of any identifiable person’s voice.
2.4 We do not use a separate third-party product analytics provider, so no Inputs or Outputs reach one.
3. What is never sent to AI sub-processors
Passwords, authentication secrets and session tokens; payment card data and billing records; your customer list or the identity of other Bidwhistle customers; and any content beyond what is needed to answer the request in hand — subject to paragraph 2.2, where the request is about a document you have uploaded.
4. Retention of AI interaction data
| Data | Held by | Retention |
|---|---|---|
| Olivia session recordings (the Authorised User’s audio and the avatar video) and transcripts | Anam | 30 days from the session, then deleted automatically; we also delete each session at Anam ourselves once it is 30 days old. Zero retention is not available on our plan |
| Olivia transcripts held in the platform | Us | We have not established whether the platform keeps its own copy of Olivia transcripts and conversation logs, or for how long. We are establishing it and will record it here. |
| Documents uploaded to the Response Workspace | No one, after the request is answered | Not stored by us, and passed through to Anthropic to answer the request. Anthropic states that it deletes API inputs and outputs within 30 days, with exceptions in its terms such as enforcing its usage policy or a legal requirement — and it commits in those terms that it “may not train models on Customer Content from Services”. |
| The usage row written for a workspace request (token counts and a reference, no content) | Us | Kept as a product and usage log — 12 months |
| Request and response bodies under roughly 10 KB on endpoints that still inherit request-history logging, which include endpoints carrying bid text and the endpoints that call the model | Us | Approximately 1 day, then overwritten. See item 9 of Annex II and Planned, not yet done: request-history logging to be disabled on the endpoints carrying customer bid text |
| Inputs and Outputs saved in your account | Us | For the life of the account. Deleted 90 days after the account ends, whether the subscription was not renewed or you closed the account |
| Voice embeddings / derived voice data | — | Not created or stored |
5. Disclosure and marking of AI-generated output
5.1 We are the provider of an AI system that interacts directly with people and generates synthetic audio and text, so Article 50 of the EU AI Act applies to us from the start of the Service.
5.2 Our website presents Olivia as an AI adviser wherever she appears, and our Terms of Service and Privacy Notice, which users accept at sign-up, say so and explain that her conversations are recorded by our voice provider and kept for up to 30 days (paragraph 2 above). Olivia does not present as a human being and never claims to be a person.
5.3 Article 50(2) of the EU AI Act requires synthetic audio and AI-generated text to carry marking in a machine-readable format so they can be detected as artificially generated. Text drafted in the Response Workspace carries the invisible watermark Anthropic adds to its model’s writing. For Olivia’s voice and video, we are confirming with our suppliers how they are marked (questions sent on 27 September 2026), and we will update this Annex and our AI Use Disclosure when we have their answer.
5.4 Flow-down. If you make Outputs available outside your organisation — publishing them, sending them to a client, embedding them in your own product — you must preserve the AI disclosure and any machine-readable marking, and you must not present an Output as human-authored or as a guarantee of any fact or outcome. The Terms of Service and the AI Use Disclosure set this out in full.
5.5 Outputs can be wrong. They are not legal, procurement, bid-writing, tax, financial or eligibility advice, and they are not a substitute for reading the original notice published by the contracting authority.
6. Human review
We do not routinely review the content of Olivia conversations, uploaded documents, Inputs or Outputs. A person at Bidwhistle may access them only where you have asked for support, where we are investigating a suspected Security Incident or a report under the Acceptable Use Policy, or where the law requires it. Access is least-privilege and logged, and never for model training or improvement. What a provider’s own staff may do with data it holds is governed by that provider’s terms, which for two of the four providers in paragraph 2 we have not yet confirmed in writing.
7. Changes
Changes to AI Sub-processors follow section 8, including the honest note in section 8.3 about shorter notice from AI providers. If a provider’s retention, training or location position changes — or is confirmed for the first time — we will update this Annex, Annex III and the Sub-processors page, notify subscribers, and you may object under section 8.4.
How to contact us
| What you need | Where to write |
|---|---|
| Data protection questions, sub-processor objections, data subject request assistance | privacy@bidwhistle.com |
| A signed copy of this DPA, audit notices, formal legal notices | legal@bidwhistle.com |
| Security vulnerabilities and incident correspondence | security@bidwhistle.com |
| Anything else | support@bidwhistle.com |
Bidwhistle OÜ, Sepapaja tn 6, 15551 Tallinn, Estonia. Registry code 17567745.
You can also complain to the Estonian Data Protection Inspectorate (Annex I(C)) or, if you are in the UK, to the Information Commissioner’s Office.
Changes to this Data Processing Agreement
We may update this Data Processing Agreement as described in section 16.2. We will not make a change that materially reduces the protection given to Customer Personal Data. Material changes are notified by email to the account owner at least 30 days in advance, and the version number and dates at the top of this document are updated. Sub-processor changes are notified under section 8 instead.
| Version | Date | Summary of changes |
|---|---|---|
| 1.0 | 31 August 2026 | First published version. |
| 1.1 | 28 September 2026 | Every supplier checked against its own published terms on 28 September 2026: the company we contract with added for each; Resend and WeWeb corrected to storage in the United States; Cartesia, Twilio and Hostinger added; each transfer mechanism recorded; the notes on analytics and the support mailbox answered, and the remaining open points narrowed. |
| 1.2 | 7 October 2026 | Customer data kept 90 days after a subscription or free trial ends (was 30, and 60 for a trial); closing the account stays 30 days. Clauses 11.2 and 11.3, Annex I(B) and the retention and portability measures. |
| 1.3 | 15 November 2026 | Two statements corrected against records already approved, and the document’s own title. (1) The UK Article 27 representative is recorded as James Rehman at privacy@bidwhistle.com, appointed 25 October 2026 — in section 14.3, in Annex I(A) and in Annex V, paragraph 1.4, all three of which read ‘to be appointed’. Privacy Notice v1.7 has published that appointment at section 18.1.2 since 4 November, so this document contradicted a published notice while itself forming part of the Agreement. (2) The signatory in Annex I(A) is named: James Rehman, sole member of the management board, as recorded in BWB-02.01. (3) The title line read ‘Bidwhistle - Privacy Notice’ and now names this document. No term, obligation, annex entry or transfer mechanism changes. The document stays DRAFT and is still not publishable: 41 bracketed open items in 9 groups remain, most of them confirmations owed by suppliers, and section 13.6.1 still cites a transfer impact assessment that has not been written. |
| 1.4 | 15 November 2026 | Made publishable, and brought into line with the Privacy Notice. (1) Deletion. Privacy Notice v1.7 keeps account and profile data and Customer Data for 90 days after the account ends, whether the subscription was not renewed or the customer closed the account, and applies 30 days only to a deletion asked for during the subscription and to purging backups. Clause 11.3 and Annex I(B) had a separate 30-day route for closing an account, which promised faster than the Service delivers and contradicted two published documents; that route is removed. (2) The 41 bracketed notes are replaced by what is true today, in public wording: where a supplier confirmation is owed it now says so in terms, and the planned measures in Annex II are marked planned rather than carrying a date nobody has set. (3) Section 9.1 no longer claims self-service deletion: access, correction and export are self-service, deletion is asked for in the Service and carried out by us within section 11. (4) Section 13.6.1 no longer cites a dated transfer impact assessment; it describes the assessment actually recorded at 13.6.2, 13.6.3 and Annexes III and VI, and says plainly that no separate document exists yet. No other term, obligation, annex entry or transfer mechanism changes. |