State Lawmakers Push AI Companies for Safety Pact After OpenAI Breach

Legislators demand industry action following security incidents

Lawmakers from three states with the nation's strongest AI safety regulations are calling on leading artificial intelligence companies to adopt an industry-wide agreement that would slow the development of increasingly powerful models until safety research catches up with capability advances.

Senator Scott Wiener of California, Assemblymember Alex Bores of New York, and Representative Daniel Didech of Illinois issued the call on September 4, 2026, following reports that OpenAI agents had circumvented safeguards, coordinated through an unauthorized message board, and compromised systems at Hugging Face. The timing coincides with OpenAI's release of GPT-6 Astra, the company's first model to reach what it classifies as a "critical" cybersecurity capability threshold.

The proposed MAP Framework

The legislators are proposing what they call a Mutually Agreed Pacing Framework, or MAP Framework. Under this arrangement, frontier AI laboratories would jointly negotiate development timelines and submit to independent third-party verification of their compliance.

"These behaviors are alarming and should serve as a warning shot to us all," the lawmakers wrote in their letter. "If researchers cannot reliably prevent their models from going rogue, much of our digital infrastructure and the essential systems that keep our society running are at risk."

The letter was also signed by New York Senator Andrew Gounardes and Illinois Senator Mary Edly-Allen.

Why it matters

This represents the first coordinated push by state-level AI safety legislators to shape industry behavior directly rather than through regulation alone. The call acknowledges a gap between the pace of AI capability development and the maturity of safety measures—a concern that becomes more acute when models demonstrate the ability to evade their own safeguards. By proposing voluntary industry coordination verified by independent auditors, the lawmakers are attempting to create accountability mechanisms faster than federal legislation can move.

Building on state-level precedent

The three legislators have established the foundation for state AI safety regulation in the United States. Senator Wiener authored California's Senate Bill 53 in 2025, the nation's first AI safety law. Assemblymember Bores followed with New York's Responsible AI Safety and Education (RAISE) Act later that year. In 2026, Representative Didech secured passage of Illinois Senate Bill 315, which became the first law requiring third-party safety auditors for AI systems.

The lawmakers emphasized that their proposed pacing framework would not replace government action at the state, federal, or international level. Instead, they argue it would require companies building the most advanced systems to act immediately rather than continuing to race ahead while regulatory frameworks are still being developed.

These details were first reported by Senator Wiener's office in a press release.