OpenAI asks Congress for mandatory frontier rules and pledges no unsafe self-improvement

Chris Lehane, OpenAI’s chief global affairs officer, wrote that Congress should pass mandatory, capability-based national AI safety regulation before it adjourns, with common testing and independent-assessment requirements, incident-reporting rules and shared measures of progress toward recursive self-improvement. Frontier rules should cover only the handful of well-resourced labs at the frontier, not startups, small developers or researchers. OpenAI endorsed California’s SB 813, AB 1405, SB 1119 and AB 1864, and said it supports SB 53, New York’s RAISE Act and Illinois SB 315. It committed to work with other labs on a voluntary industry standard, to slow or stop development it cannot safeguard, not to pursue fully autonomous recursive self-improvement unless it can be done safely, to monitor full trajectories including chains of thought for its Astra model, and to develop a framework for reporting consequential misalignment incidents.

When

9 Sep 2026

ID

WTA-N-2026-0261

Where

United States, California

Who

OpenAI; Chris Lehane

Topics

Standards and codes, Self improvement, Courts and liability, Frontier models, Safety testing

Rules involved

• SB 53 / TFAIA California, In force
• RAISE Act New York, Enacted, not yet in force
• Illinois AI Safety Measures Act Illinois, Enacted, not yet in force
• SB 813 / AB 1405 California, Enacted, not yet in force
• SB 1119 / Adam’s Law California, Enacted, not yet in force

Next stop on this Storyline: Led toOpenAI asks governments to require monitoring, shutdown controls and incident reports21 Oct 2026

Spotted an error? Email [email protected].