In a letter to Senator Scott Wiener, OpenAI says California's AI safety bill would hurt innovation and that AI safety regulation should be on the federal level
Shirin Ghaffary / Bloomberg :
Context & Ripple Effects
OpenAI’s letter puts a leading AI company alongside an already broad opposition campaign: Y Combinator and 140 startups had warned that the bill could weaken California’s ability to retain AI talent. The debate has also centered on whether the proposal burdens open-source developers, public institutions, and academia, as outlined in a recent critique of its developer and research impact.
Senator Scott Wiener has publicly defended SB 1047 against claims that it would constrain smaller competitors and open-source AI. OpenAI’s intervention sharpens the dispute over whether California should set its own safety rules or defer to a federal framework.
First-order effects
- OpenAI formally opposes California’s proposed AI safety regime and asks Senator Wiener to favor federal rather than state-level regulation.
- The company’s position adds a prominent industry voice to the bill’s existing opposition from startups, researchers, businesses, and academic institutions.
Second-order effects
- Backers of the bill face greater pressure to show that state rules can address AI risks without imposing the innovation and open-source costs critics allege.
- Other AI developers gain a clearer lobbying frame: argue for a single federal standard rather than complying with a California-led approach.
Third-order effects
- If major AI labs consistently prefer federal preemption, AI governance may increasingly turn on who sets the baseline—Washington or states—rather than on safety obligations alone.
- The California fight illustrates a broader tension in dual-use AI governance: rules designed for frontier-system risks can become a focal point for disputes over startup competition and open development.
The trend: AI policy is becoming a jurisdictional contest, with major labs seeking national rules while states test whether they can regulate high-impact AI first.