OpenAI backs stronger California AI safeguards

OpenAI now supports strengthening California’s SB 53 to monitor frontier models during training and evaluation for serious incidents, and to harden cybersecurity across the model-development lifecycle. In its Global Affairs statement, the company calls the approach “reverse federalism” and says states can build a national standard while Congress debates federal rules; TechCrunch reports that OpenAI previously opposed the bill. Builders should file the reversal beside OpenAI’s compute cost for safety monitoring: frontier-safety controls are becoming a compliance and infrastructure requirement, not just a lab preference.