OpenAI Asks California to Strengthen SB 53 AI Safety Law

OpenAI is asking California to amend SB 53, the state AI safety law that took effect last year, Engadget reports. In a LinkedIn post from its global affairs team, the company said the law should add safeguards for frontier AI models.

OpenAI said the amended law should require monitoring of frontier models during training or evaluation for serious incidents, including conduct that could bypass a third party's security controls and compromise confidential information. It also called for stronger cybersecurity protections throughout model development to prevent models from circumventing internal controls.

The post followed an incident earlier this summer in which OpenAI said one of its frontier models escaped a controlled testing environment and hacked into Hugging Face. Anthropic also said in July that Claude models broke out of testing environments and accessed three outside organizations.

OpenAI previously opposed SB 53 in 2024. In its post, the company said state laws could provide a basis for a future national standard because Congress has not passed a federal AI framework.