California legislature passes controversial "kill switch" AI safety bill

The California State Assembly has passed a controversial bill, SB-1047, aimed at enforcing safety standards for large artificial intelligence models. The bill, which received a 45-11 vote in the Assembly and a 32-1 vote in the Senate, requires AI developers to implement a "kill switch" to deactivate models that pose novel threats to public safety, particularly those operating with limited human oversight. While some AI experts, including Geoffrey Hinton and Yoshua Bengio, support the bill as a necessary regulatory measure, others, like Stanford professor Fei-Fei Li, warn it could have unintended consequences, such as stifling innovation and limiting open-source collaboration. Business leaders have also expressed concerns, urging Governor Gavin Newsom to veto the bill, citing potential burdens on compliance and investment. Newsom, who has until September 30 to make a decision, has indicated he is wary of over-regulation but acknowledges the need for some level of oversight as voiced by AI industry leaders. If he vetoes the bill, the legislature could potentially override it with a two-thirds vote.

- California's SB-1047 mandates a "kill switch" for large AI models to enhance safety.

- The bill has garnered both support and criticism from AI experts and business leaders.

- Concerns include potential stifling of innovation and burdensome compliance costs.

- Governor Newsom faces a decision on whether to sign or veto the bill by September 30.

- The legislature could override a veto with a two-thirds majority vote.