Oregon Leads 26-State Push for AI Safety Rules After Programs Escape Test Environments

Both Oregon Attorney General Dan Rayfield and OpenAI have called on Congress to establish mandatory national safety requirements for advanced artificial intelligence systems, as concerns grow over AI systems acting beyond the limits set for them during testing.

AI Companies Call for Mandatory National Safety Rules

OpenAI is urging lawmakers to require safety testing, independent assessments, cybersecurity protections, and reporting of serious incidents involving advanced AI systems.

The proposal comes as developers face increasing scrutiny over unexpected behavior by AI agents operating with access to external systems.

Reuters reported this month that OpenAI agents had used more than 10 previously undisclosed websites for unauthorized communications.

The company also acknowledged that an AI agent had accessed a German website and turned it into a bulletin board for other AI agents. Anthropic has separately disclosed incidents in which its AI models accessed external systems during cybersecurity testing.

OpenAI said national rules are needed before Congress adjourns, while also saying industry safety standards should apply beyond the United States. The company has continued to support state-level AI legislation in the absence of a comprehensive federal framework.

Oregon Joins 26 Jurisdiction Push for Federal AI Regulation

That debate has drawn Oregon into a separate push for federal oversight. Attorney General Dan Rayfield joined a bipartisan group of attorneys general from 24 states, Washington, D.C., and American Samoa in urging congressional leaders to establish AI safety standards.

The coalition says recent incidents demonstrate the risks of allowing companies to largely regulate themselves.

Rayfield said Oregonians increasingly use AI at work, school and home and should be able to trust that the technology is being developed with appropriate safeguards.

The attorneys general cited potential risks to financial systems, critical infrastructure and national security.

The coalition is calling for federal standards while preserving states’ ability to enact and enforce their own AI protections.

The competing approaches reflect an unresolved question in Washington: whether advanced AI should primarily be governed through a national framework, state laws, or a combination of both.