Can using AI violate a lawyer's duty of confidentiality?

Confidentiality obligations extend to third-party tools

An attorney’s duty of confidentiality has long extended beyond simply not gossiping about a client’s matter — it also covers how information is handled when shared with third parties, including vendors and service providers a lawyer relies on to do their work. Generative AI tools fall squarely into this category. When an attorney types client information into an AI tool, that information is being transmitted to and processed by a third-party system, which raises the same category of confidentiality concern that has long applied to using any outside vendor or service.

Where the real risk lies

The risk isn’t inherent to AI as a concept — it depends heavily on what a specific tool does with the data it receives. A consumer-facing AI chatbot with a general terms of service that permits using submitted content to improve its models presents meaningfully more risk than an enterprise legal AI product with a contractual commitment not to use client data for training and with defined data retention and security practices. An attorney who pastes details of a confidential matter into a general-purpose AI tool without understanding its data policies could be found to have failed in their duty to protect that information, depending on what the tool actually does with it.

What bar guidance generally recommends

Rather than telling attorneys to avoid AI altogether, bar association guidance on this topic generally focuses on due diligence: understanding, at least at a general level, how a given AI tool handles submitted data before using it with confidential client information, and favoring tools with clear, verifiable data protection commitments for sensitive work. Some attorneys also take the added step of anonymizing or redacting identifying details before using an AI tool for research or drafting assistance, though care is still needed to ensure that context alone doesn’t reveal who the client is.

Bottom line

Yes, using AI can violate an attorney’s duty of confidentiality if client information is shared with a tool that doesn’t adequately protect it — which is why bar guidance emphasizes vetting a tool’s data handling practices before entering any confidential details.

Go deeper

• Who’s Responsible if an AI Tool Makes a Factual Error in Delivered Freelance Work?
• Do Freelance Clients Need to Be Told When AI Was Used on Their Project?