— EU AI Act compliance

AI Transparency

Last updated: June 2026

In one sentence

Biddexis uses artificial intelligence to help you draft tender responses faster. Every output is a draft for human review. We are not a decision-maker, an evaluator, or a substitute for professional judgment.

What our AI does

Biddexis applies AI to three specific tasks:

  • 01
    Requirement extraction. Reads your uploaded tender documents and identifies mandatory requirements, evaluation criteria, deadlines, and formatting rules.
  • 02
    Draft response generation. Produces draft text for each requirement using your company profile, past projects, and team data — which you provide and control.
  • 03
    Compliance and quality analysis. Compares your drafts against the extracted requirements to flag gaps, weak responses, and missing mandatory elements.

What our AI does NOT do

  • We do not predict whether you will win a tender.
  • We do not evaluate or score your bid against other bidders.
  • We do not recommend pricing or financial commitments.
  • We do not submit bids on your behalf or interact with procurement portals.
  • We do not make any decision that affects you, your company, or third parties.

Models we use

Biddexis uses models from two AI providers, accessed via API:

Extraction & analysis — Anthropic
Claude Haiku 4.5

Identifies requirements and analyzes documents. Used because it's fast and cost-efficient for structured extraction.

Response generation — Anthropic
Claude Sonnet 4.5

Produces draft responses. Used because it generates higher-quality, more nuanced text in multiple EU languages.

Semantic search embeddings — OpenAI
text-embedding-3-small

Generates vector embeddings for the Knowledge Bank and Past Proposals semantic search features. No tender content is stored by OpenAI beyond the duration of the API call. OpenAI does not train on API customer data.

Anthropic and OpenAI are US-based companies. We have entered into Data Processing Agreements with both providers, governing data transfers under Standard Contractual Clauses and EU–US Data Privacy Framework safeguards.

EU AI Act classification

Under Regulation (EU) 2024/1689, Biddexis is classified as a limited-risk AI system. Specifically:

  • · Biddexis is a generative AI system that produces text content for human review.
  • · Biddexis is NOT a high-risk system under Annex III — it is not used for biometrics, employment decisions, credit scoring, or law enforcement.
  • · Biddexis does not engage in any prohibited practice under Article 5.
  • · All AI-generated output is clearly labeled as a draft requiring human review.

We comply with Article 50 transparency obligations: every user interacting with Biddexis is informed they are using an AI system, and every AI-generated document is labeled as such.

Human-in-the-loop guarantee

Every AI output in Biddexis requires human action before it has any effect:

  • · You upload the tender — Biddexis never sources documents on its own.
  • · You review every extracted requirement.
  • · You approve every generated response — they are saved as drafts only.
  • · You edit or regenerate any output that doesn't meet your standards.
  • · You explicitly export the final document — Biddexis never submits to procurement systems.
  • · The export disclaimer reminds you and the recipient that AI-assisted drafting was used.

Data and AI training

  • · Your tender documents and company data are used only to generate your responses.
  • · We do not train AI models on your data.
  • · Anthropic's API does not train on customer data by default — Biddexis uses these settings.
  • · You can delete all your data at any time from your account settings.

Reporting AI issues

If you encounter biased, inaccurate, or harmful AI output, please report it to ai-issues@biddexis.com. We review every report and use it to improve the system.

Updates to this disclosure

The EU AI Act becomes fully applicable on 2 August 2026. We will update this page as obligations evolve and as we adapt our system to comply with new requirements.