California SB 813 cleared Assembly Appropriations this week and now heads to a floor vote, three days after the European Commission activated enforcement of the EU AI Act. The timing sharpens a divergence that frontier developers can no longer paper over: Sacramento is building a voluntary certification regime while Brussels turns on mandatory transparency rules backed by fines of up to €15 million or 3% of global turnover.

Sen. Jerry McNerney’s bill would create a California AI Standards and Safety Commission, delegate auditing to private Independent Verification Organizations, and hand certified models an affirmative defense against personal-injury and property-damage claims. Certifications are revocable and subject to ongoing monitoring. It’s a standards-body architecture, closer to Underwriters Laboratories than to Brussels.

“It’s time for California to take the lead… since Washington, DC is unable or unwilling to do so,” McNerney said when the Senate passed the bill 31-7 in January. It cleared Assembly Privacy and Consumer Protection in June and was re-referred to Appropriations on July 1.

The EU is running the opposite play. As of August 2, chatbots must disclose they’re AI, deepfakes must be labeled, and AI-generated content must carry machine-readable marks, all policed by the EU AI Office.

The state layer is thickening around both. The Transparency Coalition’s tracker, cited by TechTimes, counts 85 AI-related laws enacted across 27 states in 2026, the highest volume on record at this point in the year. But Tech Policy Press notes only one state is poised to enact a frontier-model statute this year, following California and New York’s 2025 laws. The volume is in disclosure, procurement, and deepfake rules; the frontier question remains concentrated in a handful of capitals.

If SB 813 clears the floor, it lands on Gov. Newsom’s desk in the middle of an enforcement ramp he doesn’t control.

Sources

Sources