On October 21, 2026, Magistrate Judge Marty Fulgueira Elfenbein of the U.S. District Court for the Southern District of Florida issued a discovery order in Fontainebleau Florida Hotel, LLC v. Botach that held that work-product protection can apply to AI prompts and outputs created with AI tools, even if those documents were not generated at the direction of counsel. The case is part of a larger trend of plaintiffs seeking and obtaining AI prompts and outputs in discovery.

During damages discovery in a defamation action, the plaintiff Fontainebleau sought AI chats from the defendant Botach, as well as through subpoenas from the four providers of the AI models that were used by the defendant, arguing that they were highly relevant because they could reveal his intent, motive, and efforts to continue or amplify the alleged harm to plaintiff.

The defendant argued that the requested AI materials constituted protected work product because they included AI-assisted litigation preparation, research, draft analyses, and other materials concerning his defense.

Relying on Federal Rule of Civil Procedure 26(b)(3) and Eleventh Circuit authority interpreting this rule, the Court held that for AI inputs and outputs, work-product protection can apply, even when that use of AI was not carried out at the direction of attorneys. The Court nonetheless rejected the defendant’s blanket work-product objection, holding that protection must be shown document by document and requiring a privilege log for anything withheld.

In coming to its decision, the Court considered United States v. Heppner, 820 F. Supp. 3d 292 (S.D.N.Y. 2026), (discussed in our February 11 and February 17 posts), which denied work-product protection for materials that a represented criminal defendant generated using Claude because they were not prepared at counsel’s direction and did not reflect counsel’s strategy. The Fontainebleau Court declined to follow Heppner to the extent that decision created a categorical limitation on work-product protection based on whether counsel directed the AI use.

The Fontainebleau Court also addressed the risk that an AI provider’s production of a litigant’s AI materials directly to opposing counsel could bypass the litigant’s opportunity to identify protected work product. The Court therefore required the parties to confer about how AI material might be produced by AI providers to allow the defendant to assert any work-product claims.

Takeaways. As AI tool content proliferates, so do the efforts to obtain them in discovery. Companies should therefore consider:

• Strengthening privilege claims by using an enterprise AI tool (which does not train on inputs and maintains the confidentiality of inputs) whenever possible when using AI tools in connection with legal work.

• To the extent a client or other non-lawyer is acting at the direction of counsel when using AI tools, accurately documenting that context in the prompt (e.g., “I am doing this research at the direction of counsel for [X] litigation”).

• Ensuring that privilege logs identifying such communications clearly and accurately denote the basis for the privilege and that the AI tool was used with the expectation of confidentiality.

• Setting retention schedules to delete AI tool content after it is likely no longer needed, unless it must be preserved for regulatory or legal hold purposes.

This publication is for general information purposes only. It is not intended to provide, nor is it to be used as, a substitute for legal advice. In some jurisdictions it may be considered attorney advertising.