US Court Rules No Attorney-Client Privilege for AI Chatbot Conversations in Fraud Case

Bradley Feiner, chairman of the public company GWG Holdings, faced a federal investigation for allegedly embezzling over $150 million from company funds. Instead of consulting only his lawyers, Feiner also used Claude, an AI model by Anthropic, to analyze evidence and develop legal defense strategies. He shared information from his lawyers with Claude and received detailed documents outlining possible legal and factual arguments, which he then shared with his legal team. However, when the FBI seized his computers, they found 31 chat documents with Claude, which prosecutors sought to use as evidence. Feiner claimed these chats were protected by attorney-client privilege.

The case reached Judge Jed Rakoff of the U.S. District Court for the Southern District of New York, who issued a landmark ruling in February 2026. Rakoff decisively ruled there was no attorney-client privilege because Claude is not a lawyer and the conversations were not intended to be confidential. Users agree to Anthropic’s privacy policy allowing data collection, use for model training, and potential disclosure to third parties, including law enforcement. Therefore, Feiner had no reasonable expectation of privacy in his chats with the AI. The judge also noted Feiner did not consult Claude under his lawyers’ direction, and Claude itself stated it cannot provide legal advice.

Feiner was convicted on all charges in May 2026. The judge acknowledged AI tools could be legitimate aids if used under lawyers’ supervision, similar to interpreters or accountants. But when clients independently consult AI chatbots, those communications are not protected and may be disclosed in court. The article’s author, Israeli attorney Dr. Zvi Gabay, predicts Israeli courts would likely rule similarly, emphasizing that attorney-client privilege applies only to human lawyer-client relationships. Gabay warns the public against relying on AI as a substitute for professional legal advice, noting that while AI can assist, it does not have legal standing or confidentiality protections. He advises seeking qualified lawyers who responsibly use AI tools, comparing it to needing a license and experience to drive a car safely.