New regulatory trends in Silicon Valley: California AI bill removes liability clause, strengthens sensitive technology leak reporting mechanism.

Last year, California Governor Gavin Newsom vetoed a bill that would have required technology companies to bear responsibility for technical damages. The state legislative body subsequently launched the process of regulating artificial intelligence. On Tuesday, San Francisco Democratic State Senator Scott Wiener introduced an amendment to SB 53, proposing the establishment of the country's first AI transparency regulatory framework: companies developing artificial intelligence models with a certain threshold of computing power must publicly disclose safety assessment agreements to address potential catastrophic risks posed by the technology. According to the proposal, the companies must also submit special reports to the state attorney general in the event of major security incidents such as sensitive technology leaks, involving industry leaders including OpenAI, Alphabet Inc.'s Class C parent company Alphabet (GOOGL.US), and Anthropic. The amendment, based on the original bill submitted by Wiener earlier this year, makes key adjustments, maintaining provisions for protecting whistleblowers and developing public cloud service platforms. The platform aims to provide low-cost computing power support for startups and academic institutions. The controversial liability attribution clauses are removed. Wiener emphasized in a statement that legislation needs to strike a balance between encouraging innovation and mitigating risks: "AI technology is advancing rapidly, and policymaking must collaborate with top scientific research forces to unleash technological potential while strengthening security defenses, ultimately building a governance system that emphasizes trust, fairness, and accountability." The legislative attempt comes against the backdrop of loosened federal regulations: the U.S. Senate recently rejected a proposal to restrict states from enacting AI regulations, creating space for state legislation. It is worth noting that SB 1047, promoted by Wiener last year and passed by the state council, was strongly opposed by Silicon Valley venture capital firms and tech giants because it required companies to bear legal responsibility for technological disasters, eventually vetoed by Newsom on grounds of "excessive pressure on innovation." The provisions of the SB 53 amendment design resonate with the framework proposed by the AI Frontiers Model Working Group established after Newsom's veto, led by renowned scholars including Fei-Fei Li. Although the working group did not propose specific legislation, its technical assessment report provided professional support for the current bill. Companies such as Meta (META.US) and Alphabet Inc. have established an industry practice of regularly releasing model security guidelines. The SB 53 bill aims to elevate such practices to legal standards. The amendment is expected to enter the parliamentary committee review phase later this month, with Wiener stating that he will continue to consult with technology companies, academic institutions, and other stakeholders to improve the proposal. This years-long legislative game reflects the deep tension between the rapid development of AI technology and public safety assurance, signaling a new stage of local legislation and federal policy coordination in global technology governance.