Legal · Compliance

    EU AI Act: how Formatix is classified

    Formatix AI is used by recruitment agencies and executive search firms to turn CVs and notes into documents in their own templates. This page sets out how we have assessed Formatix under the EU Artificial Intelligence Act (Regulation (EU) 2024/1689), the safeguards in place and the dates that matter for recruitment. It is our own assessment, version of 21 October 2026, and not legal advice.

    Summary

    Our assessment
    RegulationEU Artificial Intelligence Act, Regulation (EU) 2024/1689, as amended by the Digital Omnibus on AI, Regulation (EU) 2026/1744
    Our roleProvider of an AI system (the Formatix platform), built on general-purpose AI models from Anthropic, Google and OpenAI
    CV formatting and document productionNot high-risk. Formatix lays out the facts in a candidate's CV and the consultant's notes in the firm's template. It does not filter applications, evaluate, score or rank candidates, or make or recommend hiring decisions
    Mandates longlist checkBeing assessed against Annex III, point 4, ahead of 2 December 2027 (see below)
    TransparencyWe say plainly that Formatix uses AI, which models process documents, and that every document is for the consultant to review
    Version21 October 2026. We update this page when the product or the guidance changes

    What Formatix does

    Formatix reads documents a recruitment firm already holds (a candidate's CV, interview notes, a job description) and produces the documents the firm sends its clients: candidate profiles, blind CVs, shortlist decks, trackers, assessment reports and similar, in the firm's own Word, PowerPoint or Excel template.

    • The facts come from the documents the consultant provides. Formatix is instructed to use only what they contain and not to add facts or opinions of its own.
    • Ratings, assessments and the consultant's view in a document come from the consultant's notes, never from Formatix.
    • Every output is an editable file. The consultant reviews it, changes it if needed and decides whether to send it.
    • Formatix does not decide who is shortlisted, interviewed, offered or rejected.

    Why CV formatting is not high-risk

    Annex III, point 4 of the Act lists as high-risk AI systems intended for recruitment or selection, in particular to place targeted job advertisements, to analyse and filter job applications and to evaluate candidates. Formatix's CV formatting and document production do none of these:

    • No filtering or evaluation. Formatting a CV changes its layout, not who is considered. Every CV uploaded comes back formatted; none is screened out, ranked or scored.
    • A narrow procedural task. Moving a CV's content into a firm's template is the kind of narrow procedural task Article 6(3) describes, done on a document a person has already chosen to send.
    • No profiling. Formatix does not build profiles to predict a candidate's performance, behaviour or suitability.
    • Human control. The consultant chooses the candidate, the template and the content, and approves the document before it reaches a client.

    Because Formatix is an AI system that produces text, the transparency duties of Article 50 are the relevant ones for these features, and we meet them as described below.

    The Mandates longlist check

    Mandates is a separate, optional part of the platform for running a search. When a firm uses it, Formatix checks each longlist profile against the requirements in the brief and shows the lines from the profile behind each tick, so the consultant can see why someone appears on the list. The consultant reviews every profile and makes every decision.

    Because comparing candidates against requirements is close to what Annex III, point 4 describes, we are assessing this feature on its own, ahead of the date the high-risk rules for recruitment apply (2 December 2027). We will publish the outcome on this page and make any changes it calls for before that date. CV formatting and document production do not depend on it, and firms that do not use Mandates do not use this feature.

    Transparency and safeguards

    • Disclosure. Our website, Privacy Policy and Data Processing Agreement state that Formatix uses AI models and name the providers: Anthropic (Claude), Google (Gemini) and, as a back-up, OpenAI, with Microsoft Azure for text recognition.
    • No training on customer data. We do not use uploaded documents to train AI models, and we use our AI providers under terms or settings under which they do not train on them either.
    • Instructions to the models. The models are instructed to use only the source documents, not to invent facts and not to add their own opinions about candidates.
    • Review before use. Outputs are drafts in editable formats; nothing is sent to a client or a candidate by Formatix without the firm's action.
    • Data protection. Processing is under the UK GDPR and the EU GDPR as set out in the Data Processing Agreement, with the platform hosted in London and the sub-processors and their locations listed in its Annex 3.
    • Retention. Firms can set records to delete automatically after a period they choose.

    Dates that apply

    DateWhat applies
    1 August 2024The AI Act enters into force
    2 February 2025Prohibited practices and the AI literacy duty apply
    2 August 2025Rules for general-purpose AI models apply
    2 August 2026Transparency duties (Article 50) apply
    2 December 2027High-risk rules for Annex III systems, including recruitment, apply (moved from 2 August 2026 by the Digital Omnibus on AI)

    Check the current text of the Regulation for your own obligations; dates and guidance can change.

    Documentation and contact

    We keep an internal record of this assessment: the system's functions, the models used, the safeguards and the review of the Mandates longlist check. Customers and prospective customers can ask for it, and for our Data Processing Agreement, at [email protected].

    Questions about the EU AI Act and Formatix

    Is Formatix a high-risk AI system under the EU AI Act?

    Not for CV formatting and document production. Formatix lays out the facts from a CV and the consultant's notes in the firm's template; it does not filter applications, evaluate, score or rank candidates, or make hiring decisions, which is what Annex III, point 4 covers. The optional Mandates longlist check is being assessed separately ahead of 2 December 2027.

    Does Formatix make or recommend hiring decisions?

    No. The consultant chooses the candidates, the content and whether a document is sent. Formatix produces the document.

    Does Formatix profile or score candidates?

    CV formatting and document production do not. Any rating in a document comes from the consultant's notes. In the optional Mandates feature, Formatix checks longlist profiles against the brief and shows the evidence behind each tick for the consultant to review; that feature is being assessed on its own.

    When do the AI Act's recruitment rules apply?

    The high-risk rules for recruitment and other Annex III systems apply from 2 December 2027, after the Digital Omnibus on AI (Regulation (EU) 2026/1744) moved them from 2 August 2026. Transparency duties apply from 2 August 2026.

    Which AI models does Formatix use?

    Claude models from Anthropic and Gemini from Google, with OpenAI as a back-up when the main model is unavailable, and Microsoft Azure AI Document Intelligence for text recognition. The full list of sub-processors and locations is in Annex 3 of our Data Processing Agreement.

    Is candidate data used to train AI models?

    No. We do not use uploaded documents to train AI models, and our AI providers process them under terms or settings under which they do not train on them.

    Can we get Formatix's AI Act documentation for our own compliance file?

    Yes. Email [email protected] and we will send our assessment record and the Data Processing Agreement.

    Last updated 2026-10-21.