Use AI on your legal files. Protect information before it is sent.

The main risk is sending a third-party provider information covered by professional secrecy: the client’s identity, the facts of the case, strategy, correspondence, exhibits or anything that allows indirect identification. The risk is therefore not limited to first and last names. You also need to consider the terms of the service used, any data retention and possible access. The CNB (French National Bar Council) recommends particular care over confidentiality and data security when AI tools are used in professional practice.

No. Pseudonymisation reduces risk; it is not an absolute guarantee. A document may still contain indirectly identifying elements or confidential information that does not need to name anyone to be covered by professional secrecy. The lawyer must therefore assess the content sent, the AI provider’s guarantees, whether sending is necessary and the rules that apply to the case. Pseudonymisation can be a useful layer of protection, but it replaces neither professional judgement nor other confidentiality measures.

Because generative AI can produce an answer that is plausible but inaccurate, incomplete, out of date or ill-suited to the case. It can also invent a reference or misinterpret a rule. AI must therefore remain an assistant: the lawyer stays in control of the reasoning, checks the sources and alone decides how the result is used professionally. NYMETRA acts on the protection and control of the data sent; it does not certify the legal accuracy of answers generated by the AI provider.