California Governor Gavin Newsom vetoed SB 1047 today, the year’s most contentious AI regulation bill, which would have required companies developing large AI models — those costing at least $100 million and using 10^26 FLOPS in training — to implement safety protocols against ‘critical harms.’

In his veto message, Newsom said the bill ‘does not take into account whether an AI system is deployed in high-risk environments’ and instead applies stringent standards to even basic functions, so long as they run on a large system. He announced he has signed 17 other AI bills in the last 30 days, covering deepfakes, watermarking, election misinformation, and worker protections. He also tapped ‘godmother of AI’ Fei-Fei Li, Tino Cuéllar, and Jennifer Tour Chayes to help develop ‘workable guardrails’ for GenAI.

The bill’s author, Senator Scott Wiener, called the veto a setback for public oversight of massive corporations. The bill had drawn opposition from OpenAI, Meta’s Yann LeCun, Representative Ro Khanna, and Nancy Pelosi, who called it well-intentioned but ill-informed. Anthropic had suggested amendments that were incorporated.