Many people use AI tools in their daily lives without a second thought. In fact, many of us have ChatGPT aiding our searches on our personal phones. But a federal judge in New York recently found during a criminal trial that conversations with AI tools were not protected by attorney-client privilege. This decision is believed to be a first-of-its-kind ruling that could have significant ramifications for potential waivers of attorney-client privilege.

According to prosecutors’ filings, the defendant, charged with securities and wire fraud, used the AI tool Claude to create 31 documents before his arrest that he then shared with his defense counsel. When he was arrested, the FBI executed a search warrant at his home and seized, among other things, several electronic devices. After the search, defense counsel contacted prosecutors and told them that, before his arrest, the defendant had asked Claude specific questions regarding the government’s investigation and the charges he was facing, generating dozens of documents ostensibly covering his prompts and Claude’s answers all of which were saved on his devices.  

Defense counsel further claimed that those documents were covered by attorney-client privilege. U.S. District Judge Jed Rakoff, however, disagreed and ruled that the documents were not privileged because:

1. No attorney was involved, given that Claude is not a lawyer and does not have a law license;
2. Claude’s service terms expressly disclaim that it does not provide legal advice;
3. Entering information into Claude provides a user with no reasonable expectation of confidentiality; and
4. Pre-existing documents subsequently given to one’s lawyer cannot retroactively create a privilege.

This cautionary tale will likely create a whole new chapter of what is asked for in discovery. In this case, the defendant knowingly provided an AI tool with information he had received from his lawyers and created documents based upon his prompts and that information. Given that AI tools are not lawyers, any information fed into it by a client is information shared with a third party and may constitute a waiver of the privilege related to the original, otherwise-privileged communications between lawyer and client.

Practical Guidance for Preserving Privilege

Clients should be wary about incorporating sensitive information relevant to potential future litigation that may not yet have been produced in, or subject to, discovery – including advice they have received from their lawyers – into an AI tool that is not closed and tailored for your specific organization. If you are using a more public AI source, you should be mindful about what specific information is fed into the AI tool, keeping your prompts generic and saving the actual advice-seeking prompts as questions for your lawyers.

Technical Considerations for AI Tool Selection

The judge’s decision draws a critical distinction between consumer-grade AI tools and enterprise AI platforms, which carries significant implications for privilege protection. Consumer-grade, publicly available AI tools typically retain user queries and outputs, use that data for a variety of purposes, including model training, and may disclose user information to third parties – including government authorities. These features defeat any reasonable expectation of confidentiality, a foundational element of privilege.

Enterprise AI tools designed for commercial use by organizations, by contrast, typically offer enhanced security features, data isolation, dedicated infrastructure and contractual terms that prohibit user data from being combined with other users' data, shared with third parties or used to train AI models. These platforms often also include terms providing companies with ownership rights to all outputs, creating a more defensible framework for preserving confidentiality.

Key technical safeguards to look for when selecting an AI tool include:

• Strict privacy policies that prohibit disclosure to third parties or government authorities;
• Data segregation mechanisms that isolate user inputs and outputs;
• Opt-out provisions for model training;
• Contractual confidentiality obligations; and
• Limited data retention periods or the ability to disable data retention entirely.

Recommended Practices for Organizations

Organizations should implement clear policies governing AI use in connection with legal matters.

Assume no privilege with public AI tools. A company's independent use of publicly available AI tools to research legal matters is likely not privileged, as providers of those tools are third parties who may retain inputs and disclose them to others.

Leverage enterprise AI tools with appropriate technical and contractual safeguards where possible. Carefully review vendor terms and infrastructure to ensure suitable degrees of data segregation. Where appropriate, consider law-specific AI tools that maintain confidentiality protocols.

Document attorney direction. If AI tools are to be used as part of litigation preparation or to facilitate providing legal advice, ensure that the use is at counsel's specific direction and properly documented, which may help support a claim for protection. Prompts may reflect that the work is being performed at counsel's direction.

Restrict access and limit dissemination of AI prompts and outputs to preserve confidentiality. Treat documents generated by public AI tools as potentially discoverable.

Train employees about AI risks and consider restricting the use of consumer AI tools on workplace systems, particularly when dealing with sensitive or potentially privileged information. Any employee who uses a non-enterprise, consumer version of an AI tool may be unwittingly generating discoverable records – including instances in which an employee inputs advice previously provided by a lawyer.

Opinions and conclusions in this post are solely those of the author unless otherwise indicated. The information contained in this blog is general in nature and is not offered and cannot be considered as legal advice for any particular situation. The author has provided the links referenced above for information purposes only and by doing so, does not adopt or incorporate the contents. Any federal tax advice provided in this communication is not intended or written by the author to be used, and cannot be used by the recipient, for the purpose of avoiding penalties which may be imposed on the recipient by the IRS. Please contact the author if you would like to receive written advice in a format which complies with IRS rules and may be relied upon to avoid penalties.