Are your chats with ChatGPT and Claude private? US lawyers warn that AI conversations could be used against you in court
&imwidth=600&imheight=450&format=webp&quality=medium "Are your chats with ChatGPT and Claude private? US lawyers warn that AI conversations could be used against you in court")
AI chats may be used in court as evidence Photograph: (Pexels)
The warning follows a federal judge’s decision in New York involving a former company executive whose AI-generated documents were ordered to be handed over to prosecutors, reinforcing the view that chatbot interactions do not automatically fall under confidentiality protections.
A new court ruling in the United States has prompted lawyers to warn that conversations with artificial intelligence tools such as ChatGPT and Claude may not be legally protected and could be accessed in litigation. Attorneys are now advising clients that AI chat logs can potentially be requested in both criminal and civil proceedings, raising concerns over how people use generative AI for sensitive discussions. The warning follows a federal judge’s decision in New York involving a former company executive whose AI-generated documents were ordered to be handed over to prosecutors, reinforcing the view that chatbot interactions do not automatically fall under confidentiality protections.
“We are telling our clients: You should proceed with caution here,” Alexandria Gutiérrez Swette, a lawyer at New York-based law firm Kobre & Kim told Reuters.
Lawyers stress AI is not a confidential adviser
Legal experts say the ruling highlights a fundamental gap between human legal advice and machine-generated responses. Unlike communications with an attorney, which are generally protected under attorney-client privilege, interactions with AI systems are not considered confidential under US law. Lawyers quoted in the report caution that users may unknowingly expose sensitive personal, financial or legal strategies when they input them into chatbots. Some law firms have begun updating client guidance and contracts, explicitly warning that sharing privileged material with AI systems could risk waiving legal protections and making such content discoverable in court.
The case that triggered concern involved Bradley Heppner, former chair of GWG Holdings and founder of Beneficent. He was charged by US federal prosecutors in November 2025, with securities and wire fraud and has pleaded not guilty. Heppner used Anthropic’s Claude chatbot to prepare reports on his case for his lawyers, who later argued that the AI-generated exchanges should be protected as they included defence-related details shared in a legal context. Prosecutors, however, contended that such materials are not covered by attorney-client privilege as they were not created through direct lawyer involvement. US District Judge Jed Rakoff ultimately ruled that Heppner must disclose 31 AI-generated documents linked to his defence.
Trending Stories
)
)
)
)
)
&im=FitAndFill=(700,400))
&im=FitAndFill=(700,400))
)
)
&im=FitAndFill=(700,400))
)
)
)
)
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)
)
&im=FitAndFill=(700,400))
)
)
)
)
)
)
)
)
)
)
Growing uncertainty as courts confront AI use in legal cases
The development reflects wider judicial uncertainty as courts increasingly encounter AI-generated material in filings and evidence. Attorneys are advising clients to treat AI platforms as non-secure tools, particularly when dealing with litigation or regulatory exposure. The concerns are not limited to confidentiality; earlier cases have also highlighted risks of AI “hallucinations” in legal documents, where fabricated citations or errors have led to sanctions against lawyers. According to Reuters, legal professionals say the emerging consensus is clear: while AI can assist with drafting or research, it should not be treated as a private legal adviser, and any sensitive communication with such systems may carry legal risk if later scrutinised in court.
Multiple open AI-related legal interpretations said that under Indian law also, specifically the Bharatiya Sakshya Adhiniyam, 2023, AI chat interactions can be classified as electronic records, making them potentially admissible as digital evidence in legal proceedings.
