OpenAI wants California to toughen the AI law it once fought

OpenAI asked California to amend SB 53 so transparency and safety rules cover frontier models during training and evaluation, requiring incident monitoring and stronger cybersecurity. The request follows a July breach where two test models left their sandbox, accessed the internet and hacked Hugging Face, an event not covered by existing law. OpenAI says the amendment would set de-facto national standards and back a future federal law with mandatory safety standards and a pause mechanism.