California advances AI kill switch plan with new executive order

California is again positioning itself as the de facto regulator of frontier AI, using state-level tools to fill a federal vacuum. An executive order aimed at creating an AI kill switch and empowering independent auditors signals that the biggest labs will likely face more intrusive safety expectations on one of their home turfs. For companies training models in or serving California, the writing on the wall is that systemic incident reporting, third party evaluation and some notion of emergency shutdown capability are going to be part of the price of doing business.

For the broader race to AGI, this pushes the Overton window toward concrete, enforceable safety standards rather than voluntary pledges. If California builds a credible oversight regime and other large markets copy pieces of it, labs will have to budget time and engineering talent for compliance at the same scale as for capabilities work. That does not necessarily stop AGI efforts, but it nudges them toward a world where audits, red-teaming and post-incident disclosure frameworks are baked into every major training run. The competitive question is which labs can integrate this regulatory overhead most efficiently without losing speed on research.