Three More A.I. Protective Orders
Michael D. Berman examines three recent AI protective orders addressing how confidential discovery materials may, and may not, be submitted to artificial intelligence tools.
Tag protective orders
Read MoreThree More A.I. Protective Orders
Dispute over A.I. Protective Order
In U.S. v. Thomas, defense counsel and the Government disputed proposed protective-order restrictions governing the use of AI in discovery. The court declined both to compel production of purportedly non-sensitive Rule 16 discovery and to enter the Government’s proposed protective order.
Another A.I. Protective Order – Part 10
U.S. v. Dygdon provides another example of an AI protective order governing discovery materials. The consent order bars certain AI uses and requires confidentiality, no model training, restricted third-party access, and deletion at the end of the case. The article also presents Craig Ball’s argument for proportional safeguards that protect discovery without unnecessarily restricting access to affordable AI tools.
How Does an “Attorney’s Eyes Only” Order Apply to an Unrepresented Party?
A New York federal court ordered that Attorney's Eyes Only material produced in redacted form to an unrepresented defendant must come with a detailed redaction log, so they can meaningfully decide whether to challenge the designation.
A.I. Protective Order—Again
Two recent Southern District of New York cases highlight a growing trend in discovery practice: protective orders that regulate how parties may use confidential discovery materials with AI tools. While not universal, these provisions increasingly require enterprise-grade security, prohibit model training on produced data, and establish disclosure obligations for AI platform use.
Unicorn Rejects A.I. Protective/Confidentiality Order – Order Entered in Criminal Case
In Litton v. Roblox Corporation, the Northern District of California declined to modify its model protective order to address AI use, relying on existing standing-order provisions. Michael D. Berman contrasts that decision with a growing body of civil and criminal cases imposing AI-specific discovery restrictions and examines whether AI protective-order provisions are becoming a standard of care.
A.I. Protective Orders Are Becoming Routine
Courts are increasingly entering AI protective orders that restrict how parties use generative AI with discovery materials, highlighting growing concerns over confidentiality, privilege, clawbacks, and litigation risk.
Lively v. Wayfarer Parties/Baldoni – Requests to Seal, Unseal, and a “Bottom-Line Order”
In its latest ruling, the Lively court rejected overbroad sealing requests, emphasizing public access and judicial accountability in discovery and summary judgment records.
Order Prohibiting Upload of Confidential Discovery Documents to Artificial Intelligence (“AI”)
A federal court barred uploading confidential discovery materials to AI, raising key eDiscovery and confidentiality issues for future Rule 26(f) discussions.
Local Rule Prohibiting Definitions in Interrogatories Without Leave of Court
The District of Nebraska prohibits definitions in interrogatories without prior court approval, as highlighted in Payne v. Geer. This contrasts with Maryland and other jurisdictions that provide structured, permissible definitions to streamline discovery. The debate highlights differing judicial philosophies on clarity versus burden in discovery.