AI-generated docs arent covered by attorney-client privilege, judge says

Source: Mashable Tech

Background

Manhattan-based U.S. District Judge Jed S. Rakoff addressed the status of AI‑generated documents during preliminary trial proceedings. The issue arose in the case of Beneficient CEO Bradley Heppner, who is charged with $150 million in securities and wire fraud spanning 2018–2021. Heppner was indicted in November 2023 by the U.S. Attorney’s Office for the Southern District of New York.

Before his arrest, Heppner used Anthropic’s Claude chatbot to create 31 documents, which he later shared with his defense attorney. Federal investigators seized those documents and argue they are “fair game” and should be treated as work product rather than privileged legal strategy. The prosecutors also note that the AI tool’s usage policies do not guarantee confidentiality.

Judge’s Decision

Rakoff ruled that the AI‑generated documents are admissible in court and fall outside attorney‑client privilege. He found no legal basis for treating the materials as privileged, although he acknowledged that their use could create a witness‑advocate conflict and potentially lead to a mistrial.

Reactions and Implications

The decision highlights a growing clash among AI developers, privacy watchdogs, and safety advocates.

• AI industry response: OpenAI CEO Sam Altman has advocated extending attorney‑client and therapist‑client privileges to AI chatbot conversations, arguing that increasingly personal uses—such as therapy or health advice—warrant legal protection.
• Legal challenges: AI companies face numerous lawsuits concerning copyright infringement, mental‑health impacts, and youth safety. Many developers have resisted full transparency of chat logs, even though some grant government entities permission to review them.
• Privacy concerns: watchdog groups have raised alarms about extensive data collection and storage by AI tools, prompting developers to implement measures that limit chat‑history retention and offer “incognito” modes for users.

These developments suggest that courts may increasingly confront questions about the confidentiality of AI‑generated content and the scope of existing legal privileges.