TL;DR
According to Reuters on September 10, 2026, OpenAI publicly called for mandatory AI safety requirements nationwide, endorsed four California bills (SB 813, AB 1405, SB 1119, AB 1864), and said it will continue supporting state legislation until Congress acts — a 180-degree reversal from its 2024 opposition to California's SB 1047.
Event overview
On September 10, 2026, per a Reuters report, OpenAI publicly urged mandatory AI safety requirements across the US and said it would keep supporting state legislation until Congress acts. OpenAI explicitly endorsed four California AI bills: SB 813 (independent safety evaluation infrastructure), AB 1405 (AI auditor standards), SB 1119 (minors protection), and AB 1864 (preventing AI biological threats). Ifeng Finance's AI research unit described the move as nearly a 180-degree reversal in OpenAI's regulatory stance.
Key facts
| Item | Detail |
|---|---|
| Date | September 10, 2026 (per Reuters) |
| Actor | OpenAI |
| Core position | Push for mandatory US AI safety requirements; keep supporting state legislation until Congress acts |
| Endorsed bills | SB 813 (independent safety evaluation infrastructure), AB 1405 (AI auditor standards), SB 1119 (minors protection), AB 1864 (preventing AI biological threats) |
| Stance shift | Opposed California SB 1047 in 2024, arguing third-party safety evaluations and incident reporting would stifle innovation; favored federal primacy and state preemption in 2025 |
Background: California passed 26 AI bills in its current session, 24 of which await the governor's signature; SB 813 passed the state Senate 37-0. The US House has passed a ten-year freeze on state AI regulation. OpenAI had previously backed California SB 53 (Frontier AI Transparency Act) alongside Anthropic and others; the law took effect January 1, 2026, requiring frontier model developers to publish safety frameworks and report major incidents.
Background and impact
The shift comes amid intensifying regulatory competition between federal and state levels: the White House favors a light-touch path that would preempt state AI laws, while California — home to most tech giants — effectively sets the national floor. OpenAI is pursuing a dual-track strategy: uniform rules if Congress legislates, or at minimum support for state laws of its choosing if it does not. As of September 10, 2026, no formal federal AI safety bill has been enacted.