A federal judge just established a precedent that every AI tool user should know about: nothing you type into an AI chatbot is legally privileged. Not even close.
In United States v. Heppner, decided February 17, 2026, a Southern District of New York court ruled that a defendant's conversations with Anthropic's Claude could be used as evidence. The defendant argued those chats should be protected under attorney-client privilege and work product doctrine. The court rejected both arguments on every count.
The Three Strikes Against Privilege
The ruling dismantled the privilege claim in three parts. First, no attorney-client relationship can exist with an AI. The court noted that attorney-client privilege requires "a trusting human relationship" - something an AI model simply cannot form. Second, the conversations weren't confidential. AI platforms retain user data for their own business purposes, including model training, which means a third party has access to the content. Third, the defendant was acting independently, not seeking actual legal advice from a lawyer through the tool. Claude itself disclaims the ability to provide legal advice.
The work product doctrine failed too. That protection applies to materials prepared "by or at the behest of counsel." Typing questions into a chatbot on your own doesn't qualify.
What This Actually Means for You
This ruling has teeth beyond criminal cases. If you're a freelancer discussing client contracts with ChatGPT, a small business owner running legal questions past Claude, or a marketer sharing campaign data with any AI assistant, that information is potentially discoverable in court. There is no privilege protecting it.
The practical gap here is significant. Millions of people treat AI chatbots like confidential advisors. They paste in contracts, employee disputes, financial details, and legal questions. This ruling confirms what privacy advocates have warned about: AI platforms are third parties, and sharing information with them destroys any expectation of confidentiality.
This also matters for businesses with employees using AI tools. If a team member pastes proprietary information, client data, or sensitive legal communications into an AI chatbot, that information could surface in litigation or regulatory investigations. The "I was just asking AI for help" defense doesn't hold up.
Practical Steps to Take Now
The National Law Review, which published the underlying analysis, recommends several concrete steps:
- Set clear AI usage policies for your team. Specify exactly what categories of information cannot go into AI prompts - client names, financial figures, legal strategy, trade secrets.
- Train employees on what acceptable AI usage looks like. Provide example prompts that get useful results without exposing sensitive data.
- Treat every AI interaction as public. If you wouldn't put it in an email to a stranger, don't put it in a chatbot.
- Use AI tools with enterprise agreements when possible. Some enterprise tiers explicitly exclude your data from training, though even that may not restore legal privilege.
The timing matters. The commercial real estate sector alone has roughly $1 trillion in debt maturing within the next year, creating exactly the kind of high-pressure environment where people cut corners and ask AI for quick legal guidance instead of calling their lawyer.
This ruling won't be the last. But it draws a clear line: AI chatbots are tools, not counselors, and the law treats them accordingly.