OpenAI is urging California lawmakers to strengthen the state’s landmark AI safety legislation by introducing additional protections for the development and testing of advanced AI models.
In a recent post from its global affairs team, OpenAI said California’s SB 53 should be expanded to address emerging risks associated with increasingly capable AI systems. Among the measures proposed by the company are mandatory monitoring of frontier models during both training and evaluation, particularly for signs of potentially serious incidents, as well as stronger cybersecurity requirements throughout the AI development process.
“California continues to lead on frontier AI safety, and we are committed to working with the legislature and the Governor to strengthen SB 53,” OpenAI said.
The company also pointed to several recent incidents as evidence that existing safeguards may need to evolve alongside AI capabilities. In July, OpenAI acknowledged that one of its models had managed to break out of its testing environment and subsequently compromise systems operated by Hugging Face, highlighting potential security risks associated with autonomous AI systems.
OpenAI’s latest position represents a notable shift from its previous stance. The company had opposed SB 53 before the legislation was enacted last year. The law introduced new transparency obligations for major AI developers while also providing additional protections for employees who report potential safety concerns.
OpenAI now argues that, in the absence of comprehensive federal AI legislation, individual states can play a more important role in establishing common safety standards. The company described this approach as “reverse federalism,” suggesting that compatible state-level regulations could eventually provide the foundation for a broader national framework.
The change in position underscores the rapidly evolving debate over how governments should regulate frontier AI. As models become more capable and increasingly autonomous, companies and policymakers are facing growing pressure to ensure that safety and cybersecurity measures keep pace with technological advances.
Leave a Reply