OpenAI calls for binding US AI safety rules

OpenAI is calling for binding US safety rules for the most powerful artificial intelligence systems, marking a shift for a company that had resisted some regulation.

Chris Lehane, OpenAI's chief global affairs officer, made the call on September 9 as US lawmakers discuss possible AI rules. OpenAI said in a policy statement that voluntary commitments by AI developers are no longer enough and urged Congress to act before it adjourns in December.

The ChatGPT maker said it wants a federal framework with common testing standards, independent assessment of frontier models, tougher cybersecurity requirements and mandatory reporting of serious safety incidents.

“The prospect of AI-accelerated AI development demands more than voluntary commitments,” Lehane wrote. He argued that US rules should be based on what models can do, rather than on company size.

OpenAI said it was now backing some measures it had previously declined to endorse. The company endorsed four California state bills, including some it had lobbied against in the past. OpenAI had opposed US states setting their own AI rules and had asked Congress to create federal laws.

The shift comes a week after OpenAI released GPT-6 Astra, which it has described as its most capable model to date. OpenAI President Greg Brockman characterized the release as the beginning of the era of artificial general intelligence, the goal of creating AI that equals human abilities in most fields.

The new position brings OpenAI closer to Anthropic's support for tighter regulation. Critics close to the White House argue that large AI companies are seeking rules that would protect their dominance and raise costs for smaller competitors.

OpenAI said any national safety framework should apply only to the handful of companies building the most powerful models and should not be used to restrict open models, whose underlying technology is released publicly for developers to download and adapt.