Legal AI saves time only if it can cite the actual texts
In brief
According to Journal du Net, notaries, lawyers and legal staff already use ChatGPT, Claude or Mistral, often without IT approval. Answers look fluent, but references to the texts stay vague, so the professional has to redo the research. The article says what now matters is not the name of the model, but the texts the system can actually reach, in which version, and with which links.
Source : Journal du Net - « L'IA sait raisonner sur le droit. Elle ne sait pas où sont les textes. »
What it changes for an SME
If your firm of 10 to 250 people uses generative AI for legal work, a well-written answer without a checkable source does not cut your time or your liability. According to Journal du Net, drafting time is lost again when staff must find the applicable article, its current version and related decisions. Buying a second specialised legal tool can mean paying twice, on top of the general assistant already in use. Connecting that existing assistant to indexed search over legal texts, without sending client names into the search, is presented as a simpler way to handle risk and confidentiality.
What you can do
- Check whether staff already use ChatGPT, Claude or Mistral for legal questions outside your internal rules.
- Do not rely on an AI legal answer unless it cites a dated, verifiable text that is currently in force.
- Ask whether your current assistant can connect to a legal corpus, for example through an open connector such as MCP, before adding another tool.
Let's talk about the time you could save.
In 20 minutes, we find where AI could save you time or money. Free, no commitment.