SB 1047

SB 1047, introduced by California State Senator Scott Wiener in February 2024, would have required companies training the largest AI models (those costing more than $100 million or using more than 10^26 FLOPS of compute) to perform safety assessments before release, implement kill-switch capabilities, and provide whistleblower protections for employees identifying safety risks. The bill passed the California legislature but was vetoed by Governor Gavin Newsom in September 2024. It was endorsed by Lawrence Lessig, Yoshua Bengio, Geoffrey Hinton, Stuart Russell, and others as 'the bare minimum for effective regulation of this technology.' Lessig's endorsement in a Boston Globe op-ed became a touchstone of the public debate. The bill's defeat — and the subsequent introduction of weaker successor legislation — is the paradigmatic recent case of the legislative modality of AI governance encountering the political power of the industry it attempts to govern.