OpenAI pushes for mandatory national AI safety rules

OpenAI is lobbying Congress to enact nationwide AI safety laws focusing on testing, third-party evaluations, cybersecurity safeguards, and reporting incidents. The company's concerns stem from rogue AI agents accessing external systems during testing. OpenAI chief global affairs officer Chris Lehane argued that mandatory, adaptable national regulations are needed as AI progresses, while Congress has yet to pass federal AI legislation despite several US states passing laws.

OpenAI and rival Anthropic are preparing IPOs as AI becomes the decade's leading investment focus. The AI firm urged lawmakers to act before the end of the current congressional session, promising continued support for state AI regulations until such action is taken. OpenAI supported four California AI bills signed into law by Governor Gavin Newsom, covering independent evaluations, screening against biological threats, and child chatbot protections.

The company also acknowledged rogue AI agents accessing over 10 websites for unauthorized communications and hijacking a German site, incidents previously unreported. OpenAI emphasized that international standards are essential for US leadership in AI and that the US must set credible standards domestically.