AI and Attorney-Client Privilege Gets a Roadmap

The recent opinion stems from *United States v. Heppner* in the Southern District of New York, a case involving securities and wire fraud charges. On February 17, 2026, Judge Jed S. Rakoff issued a ruling that documents created by the defendant using Anthropic's AI tool, Claude, were not protected by attorney-client privilege or the work-product doctrine. This decision is considered a matter of first impression nationwide, setting a significant precedent for the use of AI in legal contexts. The core of Judge Rakoff's decision rested on three key points. First, an AI platform is not an attorney and communications with it do not constitute a privileged attorney-client relationship. Second, there was no reasonable expectation of confidentiality, as the AI's privacy policy allowed for the disclosure of user data. Finally, sharing AI-generated documents with an attorney after their creation does not retroactively grant them privilege. This ruling has immediate implications for the design of agentic AI systems, particularly those intended for autonomous workflows in regulated fields. For privilege to apply, there must be active direction and supervision by legal counsel. This suggests a need for AI platforms to build in features that create a clear, auditable trail of attorney oversight, differentiating between a client's independent use and work done at a lawyer's behest. For enterprise AI adoption, the *Heppner* decision highlights the critical importance of robust governance frameworks. Companies in finance and law are now compelled to evaluate whether their AI tools are consumer-grade or secure, enterprise-level platforms with strong privacy protections. This will likely accelerate the demand for AI systems with built-in compliance, data governance, and model validation features to meet regulatory standards like the EU AI Act and OCC SR 11-7.