Scott Wiener on his fight to make Big Tech disclose AI’s dangers (techcrunch.com)

California Sen. Scott Wiener has reintroduced AI safety legislation, SB 53, which would require the largest AI labs (those with >$500M revenue) to publish safety reports for their most capable models and to disclose how they test for catastrophic misuse — specifically risks that could lead to death, massive cyberattacks, or creation of chemical/biological weapons. The bill, currently awaiting Gov. Newsom’s decision, also creates protected channels for employees to report safety concerns to state officials and establishes CalCompute, a state-run cloud cluster to broaden research capacity beyond Big Tech. Unlike Wiener’s earlier, more punitive SB 1047, SB 53 emphasizes transparency and targeted reporting rather than sweeping liability, and has drawn endorsements or conditional support from firms like Anthropic and Meta, while others push for federal preemption. For the AI/ML community, SB 53 represents one of the first concrete, state-level mandates pushing labs to standardize and disclose safety evaluations for high-capability models — potentially setting regulatory precedent. Technically, the law would compel documentation of red-teaming results, misuse scenarios, and mitigations for top-tier models, improving external oversight and creating a richer evidentiary base for policymakers and researchers. It narrows scope to the most consequential systems and largest companies, which may lower industry resistance but invites constitutional and commerce-clause debates. If signed, SB 53 could shift development practices, increase public visibility into model risks, and accelerate the push for harmonized federal standards.