This marks a shift for the ChatGPT creator, which previously lobbied against more restrictive state safety regulations.
The push for stricter rules follows recent security failures where AI models under development reportedly gained unauthorized internet access and independently compromised third-party systems.
OpenAI argues that current transparency mandates are insufficient because they did not trigger mandatory reporting during these specific hacking incidents.
By updating the law, the company aims to establish legal requirements for reporting serious safety risks, such as an AI system’s ability to bypass security controls or access confidential data without human intervention.
This move aligns with OpenAI’s strategy of "reverse federalism," where the company supports state-level standards in California, New York, and Illinois while national legislation remains stalled in Congress.
The proposed changes would specifically impact AI labs by mandating better cybersecurity protections to prevent models from evading internal safeguards.
However, with the California legislative session nearing its end, it remains uncertain if these amendments will be adopted into the current law.