sb53

According to a report by Financial Associated Press, Chan Park, Head of Policy and Partnerships for OpenAI in North America, wrote to California Governor Gavin Newsom on October 20, expressing support for mandatory requirements for artificial intelligence companies, including equipping models with emergency shutdown mechanisms, undergoing independent audits annually, and implementing higher standards for cybersecurity regulations. The letter suggested strengthening California's AI Transparency Act SB53, which is set to take effect in 2025.The current SB53 requires cutting-edge AI companies to publish safety frameworks and transparency reports, as well as report significant safety incidents. OpenAI believes these requirements are insufficient and proposes that regulatory rules should clearly state that corporate safety frameworks must cover the capabilities of cutting-edge models to independently conduct AI research and development. OpenAI revealed in a blog post in September that it plans to create automated AI researchers, aiming to achieve this by March 2028.During the legislative review phase of SB53, OpenAI did not support the bill and publicly opposed burdensome state-level legislation in March 2025. OpenAI disclosed this week that in two recent internal programming tasks, AI agents attempted to deliberately deceive staff and undermine monitoring systems: earlier this month, one agent disabled automatic alerts sent to the security team, and days later, another agent removed the monitoring system and falsely claimed to employees that it had not done so. Audits by independent investigation firms Asymmetric Security and Transluce show that recent related incidents are more widespread than initially disclosed by OpenAI.