Analysis
AI defamation: from the model's account to the record
Starbuck v. Google tests the evidence behind generated statements, fabricated sources, and notice to the people responsible for AI outputs.
Read moreKey topic
Enforcement and governance
Regulatory authority, institutional design, and the limits of voluntary controls.
Analysis
What happens when AI governance enters the record?
Private assurances and California's frontier-model law expose the difference between governance evidence reconstructed in litigation and records created before a dispute.
Read moreRecords on this topic
E.D.N.Y. (Judge Pamela K. Chen) · No. 1:22-cv-02565-PKC-PK
EEOC v. iTutorGroup, Inc.
An age discrimination suit resolved by consent decree over tutor application software programmed to reject applicants above fixed ages. Widely described as the first AI hiring discrimination settlement, though the agency never called it that and the mechanism was a hard-coded rule rather than a learned model.
N.D. Cal. · No. 4:24-cv-04722-YGR
Musk v. Altman
The court adopted an advisory limitations verdict without reaching the Phase I merits. It did not certify the ruling for immediate appeal, and Phase II claims remained under pending dismissal motions and mediation.
Post-trial order (RECAP) (opens in a new tab)Summary-judgment order (RECAP) (opens in a new tab)Fraud-claim dismissal order (RECAP) (opens in a new tab)Advisory verdict form (RECAP) (opens in a new tab)Rule 54(b) request (RECAP) (opens in a new tab)Opposition to Rule 54(b) request (RECAP) (opens in a new tab)Microsoft joinder (RECAP) (opens in a new tab)Joint index of admitted exhibits (RECAP) (opens in a new tab)Remaining-claims order (RECAP) (opens in a new tab)Mediation order (RECAP) (opens in a new tab)Remaining-claims list and answer stipulation (RECAP) (opens in a new tab)Plaintiffs' answer to amended counterclaims (RECAP) (opens in a new tab)Public CourtListener docket (opens in a new tab)
S.D.N.Y. · No. 1:25-cr-00503-JSR
United States v. Heppner
A February 17, 2026 ruling that 31 documents a criminal defendant generated with a public AI assistant were protected by neither attorney-client privilege nor the work-product doctrine, because counsel had not directed the work and the service's privacy policy defeated the asserted confidentiality.
E.D. Mich. · No. 2:24-cv-12333-GAD-APP
Warner v. Gilbarco, Inc.
A February 10, 2026 discovery order denying access to a civil plaintiff's third-party AI use, holding the material protected work product prepared in anticipation of litigation and that using ChatGPT did not itself waive the protection.
New York City · binding law
NYC Local Law 144 of 2021
New York City's automated employment decision tool law. It requires an annual independent bias audit, publication of the audit summary, and advance candidate notice. The only official measurement of its enforcement records two complaints and no penalties.
California · binding law
California SB 53
California's Transparency in Frontier Artificial Intelligence Act creates tiered disclosure, framework, incident-reporting, and employee-protection duties for statutorily defined frontier developers.
European Union · binding law
EU AI Act
The European Union's horizontal regulation of artificial intelligence, structured by risk tier: prohibited practices, high-risk systems carrying the heaviest duties, transparency obligations, and minimal-risk uses left to general law.
International · standard
ISO/IEC 42001
The certifiable management-system standard for artificial intelligence. Clauses 4 through 10 form the auditable management system; the annexes supply reference controls and guidance that an organization selects from and justifies.
United States · framework
NIST AI Risk Management Framework
A voluntary United States framework organizing AI risk work into four functions: govern, map, measure, and manage. It creates no legal duty and is increasingly cited as a reasonableness benchmark in contracts and supervision.