Our AI charter for trustworthy legal AI

AI raises the bar for reliability. It never replaces it.
Approved by you
Sourced
Sovereign
These are the commitments that govern Cyrius, the AI built into Axiocap, and the MCP connector.
Axiocap AI charter: trustworthy legal AI

Our conviction

💡 In-house lawyers, attorneys and accountants have already embraced AI, and that is a good thing. The real risk is not AI itself, it is using it without safeguards.

  • Minutes pasted into a tool beyond anyone's control means confidential data, a company's or a client's, leaving every safeguard behind.
  • As a software provider, our responsibility is to offer those safeguards: permissions, traceability and no reuse of data.
  • Our core role does not change: remaining the source of truth for your governance data.
    Responsible use of AI by legal professionals

    What Axiocap's AI does, and what it never does

    what it does

    Axiocap's AI:

    • Answers from your Axiocap data only
    • Cites the document and paragraph behind its answer
    • Prepares operations and submits them to you
    • Works with the model of your choice
    What Axiocap’s AI does
    what it never does

    Under no circumstances can it:

    • Record a sensitive entry without your approval
    • Open data the question does not call for
    • Use your data to train a model
    • Make a decision in your place
    What Axiocap’s AI never does

    Hosting, compliance and choice of model

    The technical safeguards behind Axiocap's AI
    Hosted in France

    Your data and documents are hosted in mainland France by French providers (OVH, Orange).

    They are processed in an isolated environment dedicated to each client.

    GDPR and security

    GDPR compliance, no transfers outside Europe and a dedicated DPO.

    Secure access with two-factor authentication and SSO, and full reversibility: you can take your data back whenever you want.

    The model of your choice

    Cyrius works with Claude, ChatGPT or Mistral, and even with an open-source model hosted on your own infrastructure.

    A model provider can cut off access overnight: depending on no one is part of sovereignty.

    AI charter FAQ

    Is my data used to train Axiocap's AI?

    No, never. Neither your data nor your clients' data is used to train an AI model, and it is never reused for any other purpose.

    Where does Axiocap's AI process my data?

    In mainland France, with French hosting providers (OVH, Orange), in line with the GDPR, with a dedicated DPO and no transfers outside Europe.

    How does Axiocap's AI limit hallucinations?

    The AI only opens the data relevant to your question, and every answer cites its source, down to the document and paragraph, so you can check it. Nicolas Surleau, Axiocap's co-founder, explains this in his interview on AI and data sovereignty.

    Can Axiocap's AI act without my approval?

    No. It searches, prepares and suggests, but no sensitive change, such as a share transfer or a register update, is recorded without your approval. This applies to Cyrius and to the MCP connector alike.

    Can I choose the AI model used by Cyrius?

    Yes. Cyrius works with Claude, ChatGPT or Mistral, and even with an open-source model hosted on your own infrastructure.

    Data comes before AI

    AI is only as good as the data it reads. Axiocap structures and maintains your governance data: company records pre-filled from the French official registries (Infogreffe, INPI), legally enforceable statutory registers and a time-stamped, authenticated history.
    Reliable governance data, the foundation of Axiocap’s AI