In United States v. Heppner, a Southern District of New York judge ruled that communications with a public AI platform were not privileged because the tool itself is not an attorney, its terms of service did not guarantee confidentiality, and sharing the material with counsel afterward did not retroactively create protection. Other courts have taken a narrower view. In Warner v. Gilbarco, Inc. and Morgan v. V2X Inc., judges treated AI tools as instruments rather than third parties, allowing work-product protection for AI-assisted litigation materials under Rule 26(b)(3) in appropriate circumstances. The distinction turns on whether the platform is public or proprietary, whether it retains user data, and whether contractual terms restrict training use and onward disclosure.
The critical uncertainty is how courts will treat the full spectrum of AI use—from ChatGPT to closed enterprise systems—as litigation progresses. Privilege analysis remains fact-specific and platform-dependent, with no uniform standard yet established.
Attorneys must now treat AI prompts and outputs as potentially discoverable electronically stored information subject to preservation obligations once litigation is reasonably anticipated. Corporate legal teams should incorporate AI use into litigation hold notices, eDiscovery protocols, and protective orders from the outset. The safest approach is to assume AI-generated materials lack privilege protection unless created through a closed system with contractual confidentiality safeguards and used under attorney direction for legal advice.