Courts begin shielding litigation AI chats from discovery

Early rulings in New York and Texas treat some prompts and outputs as protected work product while preserving key limits.

Courts in New York and Texas have issued early rulings protecting some generative AI prompts and outputs from discovery in litigation.

Both decisions found that AI-assisted case preparation can qualify as protected work product when it contains legal strategy or material created in anticipation of litigation.

In Assini v. Hayward, a New York court quashed subpoenas seeking a self-represented litigant’s prompts, uploads and outputs.

The court said using a commercial AI service did not automatically waive protection. It also warned that fabricated citations and unverified filings remain subject to sanctions.

In Tate Group Automotive v. Legacy Automotive Capital, a Texas court protected ChatGPT conversations created by a company principal.

The Texas court still ordered disclosure of discovery documents and other materials uploaded to ChatGPT, including documents covered by a protective order.

The rulings suggest that the purpose and content of an AI session matter more than the technology itself.

Protection is not uniform. A federal decision applied a narrower standard, while a Connecticut court ordered disclosure of prompts used by a testifying expert as part of the expert’s methodology.

Lawyers and clients should separate litigation-focused AI work, use strong confidentiality settings and avoid uploading protected opposing-party documents.

The rulings remain subject to further review, but they provide an early framework for handling AI-assisted legal work in discovery.

Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.