AI Disclosure Bill Lands on Newsom's Desk

A California bill imposing an AI-disclosure duty on the State Bar is on the Governor's desk, and the clock is short. AB 1651 (Dixon) — "State Bar of California: artificial intelligence" — was enrolled and presented to Governor Newsom on 10 August, which starts a 12-day gubernatorial window under the state constitution and puts the decision on or about 22 August. Beginning 1 January 2028, it would require the State Bar to disclose any AI-generated content used to develop or administer the bar exam and study materials — and, notably, that disclosure duty applies even where a person later revised or reviewed the AI output. The bill cleared both houses without dissent (Assembly 68-0, Senate 36-0); if Newsom neither signs nor vetoes, it becomes law anyway.

Sources: Assembly Daily History — 10 Aug 2026 · Senate Judiciary analysis — AB 1651

Six California AI bills reached third reading — none has passed the floor yet. On 17 August, five tracked measures were ordered to third reading: SB 867 and SB 1119 (Padilla — companion chatbots and children's safety), AB 686 (Berman — election deepfakes), AB 1609 (Zbur — customer-service chatbots), and AB 2575 (Ortega — health-care AI). On 18 August, AB 2504 (Bauer-Kahan — community-college AI pilot) went to the special consent calendar, making it the furthest along. Reaching third reading is position, not passage — no AI bill has yet cleared a second-house floor vote. Confirm any floor amendments to SB 867 are adopted by the 21 August cliff, then track passage votes through month-end.

Two EU AI Act deadlines now share one date: 2 December. The Article 50(2) transition for synthetic-content systems placed on the market before 2 August 2026 lands on 2 December 2026 (105 days) — the same day technical safeguards attach to the prohibition on AI-generated CSAM and non-consensual intimate imagery. Confirm your machine-readable marking plan covers every legacy generative product still in market.

NIST's public-AI-documentation draft closes for comment 16 September. The initial public draft of NIST AI 300-1, covering public-facing AI documentation guidance and templates, is open for comment for 28 more days. Because US control frameworks and a growing number of state statutes borrow NIST language, this draft is likely to become the de facto reference for "adequate" public AI documentation. File comments before 16 September and map the templates against your current model documentation.

If you build or score bar exams, licensing tests, or any regulated assessment, read AB 1651's disclosure trigger now — it survives human review of the AI output, which defeats the usual "a person checked it, so it isn't AI-generated" position. Then flag every professional-licensing body you advise: this is the drafting pattern most likely to be copied.

Thursday brings our weekly deep-dive edition — a longer read across the week's regulatory movements.

CyberEyeQ — Actionable Regulatory Intelligence
Questions or feedback: [email protected]