DraftVersion 0.1 · 2026-11-03

AI and Data Sources Notice (Tender CRM)

DraftDraft v0.1 dated 2026-11-03. This text has not been reviewed by a lawyer and may change; values in [square brackets] are not yet defined. The Ukrainian version is the master text. Current prices and plan terms are on the Pricing page and take precedence over any mention of cost in these drafts.

ItemValue
Version[VERSION]
Effective date[DATE]
Forms part ofSupplementary Documents of the Public Offer (clause 1.9, sections 11–12)

1. Purpose of this notice

It briefly and transparently explains how artificial-intelligence (AI) features work in Tender CRM, what data is transferred and to whom, where tender data comes from and what you must verify yourself. It supplements the Offer, the Privacy Policy and the DPA.


2. Which features use AI

  • a chat assistant for a tender (answers with references to documents), pre-defined queries (requirements register, goods analysis, qualification risks, deadlines and payment, bid-submission checklist, summary of changes);
  • extraction and classification of information from documents (requirements, data, line items);
  • search by meaning (indexing documents as vector representations — embeddings);
  • proposals of actions in the Service (change stage, create task, add note, etc.) — only after your confirmation;
  • summaries, tables and drafts (artifacts).

3. What data is transferred and to whom

3.1. To process a request, the following is sent to an external AI vendor: text (excerpts) of tender documents (including content extracted by OCR), texts of questions/answers from the platform, core tender fields (title, status, deadlines, stage, responsible person, recent changes), your query and files you uploaded to the chat.

3.2. Not sent: data of other Customers; credentials for platforms and keys; payments and financial records; e-mails and phones of Service users; "raw" original responses of platforms not selected for the request. Files uploaded to the chat have phones, e-mails and bank cards masked.

3.3. AI vendors (sub-processors): [OpenAI — used for chat, extraction and embeddings]; [Anthropic — optional, as an alternative chat vendor]. The current list, processing location and transfer safeguards are in the Privacy Policy (section 12) and the DPA (Annex 3). The embeddings vendor is currently OpenAI regardless of the choice of chat model.

3.3.1. The Customer may (by settings) use its own API keys of vendors [if such a feature exists]; the relationship with the vendor is then governed by the terms the Customer has accepted.

3.4. Tender documents are usually public, but files you upload to AI analysis may not be. Do not upload to AI analysis information that you have no right to disclose to third parties (trade secrets, confidential contracts, personal data, special categories, restricted-access information) — a reminder is placed under the upload button in the interface. The Customer is responsible for what it uploads.


4. Use of data for training models

4.1. Requirement on the Provider and vendors: Customer Content must not be used to train or improve third-party models, and the Provider does not use it to train its own models.

4.2. Verification status: this requirement is ensured by the terms (DPA/API terms) of the relevant vendors; the Provider verifies the vendors' current terms before every change of vendor/tier and periodically [date of last verification: [DATE]; sources: [LINKS]]. Vendors' terms may change; we will notify of a material change. [Do not publish as a final statement until verification is documented.]

4.3. Vendors may temporarily store requests/responses to prevent abuse within their terms (e.g. up to [30] days) [verify for each vendor].


5. Where tender data comes from

5.1. The Service aggregates public data from electronic platforms and registers, including: Prozorro, Zakupivli, e-Zamówienia / BZP, TED, Baza Konkurencyjności, and from Sources configured by the Customer (e.g. e-mail inboxes). The Provider is not affiliated with these platforms and is not their official representative.

5.2. We do not guarantee completeness, accuracy, currency and timeliness of this data. Delays, gaps and errors are possible (depending on the platform, its API and rate limits). Data from Sources is subject to their own terms. For some sources certain data (status, questions and answers, results, prices) may be unavailable due to the source's limitations.

5.3. Only the original platform and the contracting authority's documentation are official. Verify all critical data (deadlines, terms, requirements, price parameters, procedure status) on the platform before submitting a bid.

5.4. If you provided the Service with credentials for a platform (e.g. Zakupivli), they are stored encrypted and used only for functions you initiate.


6. Limits of AI and your duty to verify

6.1. AI outputs may be inaccurate, incomplete, outdated, fabricated (including quotes, pages, amounts, dates), contradictory or insensitive to context. They may contain bias.

6.2. Outputs are not legal, financial, tax or procurement advice. The Service does not guarantee winning a tender or acceptance of a bid.

6.3. You must independently verify outputs against primary sources (contracting authority's documentation, the platform, legislation) before using them in a bid, correspondence or decisions. Answers contain references to document excerpts — check them. If information is not in the documents, the assistant should say so rather than guess, but this cannot be guaranteed.

6.4. Human oversight: the Service does not make independent decisions with legal consequences for you or third parties. Any change in the Service proposed by AI is performed only after your explicit confirmation on a card describing the exact change.

6.5. Prompt injection. Third-party documents may contain malicious instructions. We treat their content as data, not commands, and apply technical safeguards; however the risk cannot be completely excluded. Do not trust an answer without verification, especially if it proposes unusual actions.


7. Transparency for the EU

You are interacting with an AI system; answers are generated automatically. [Add labelling in line with EU AI Act requirements — after legal review.] We do not perform emotion recognition, biometric categorisation or social scoring.


8. Questions

r.latypov@plenergy.co.uk / r.latypov@plenergy.co.uk.