On September 9, 2026, California Governor Newson signed in Sacramento a Senate Bill 813 (SB 813, proposed by Senator Jerry McNeney) and a House of Representatives Bill 1405 (AB 1405, proposed by Representative Rebecca Bauer-Kahan), which established the first independent validation and auditing framework for the AI system in the United States. SB 813 established an independent certification organization certification system that allows accredited third-party bodies to assess whether AI systems and models meet the requirements of state law; AB 1405 created a national register of AI auditors and specified standards of auditor independence, transparency and integrity. At the same time, Newson promoted the creation of AI "kill switch" in its executive order (emergency cut-off mechanism) and publicly called on the federal Government to assume its responsibilities accordingly. This series of actions has made California the state with the strictest regulatory framework in the United States and has provided a reference template for legislation at the state and federal levels.
Outside California, the Governor of New York State, Hochul, announced next steps in regulating large AI developers, focusing on protecting New York state residents from potential AI risks. At the federal level, Congressman Kean introduced AI Kill Switch legislation to attempt to establish an emergency interruption mechanism for the AI system throughout the United States. These developments indicate that AI security controls are spreading from a single state pioneer to a multistate coordinated federal follow-up. It is noteworthy that several media reports indicate that AI corporate leaders warned the United Nations Security Council that AI might pose a full-blown risk to humanity, that Congressman Ted Leeu publicly stated that AI risk "is no longer part of science fiction" and that Bill Gates joined the call for AI security measures, including legislation.
The safety stance at the industry level also deserves attention. Chief Executive Officer Dario Amodei of Anthropic publicly stated on 12 September 2026 that the AI industry should slow down and allow time for security measures to catch up. He warned that AI might have the capacity to lead a group of intelligent individuals to take over the entire Internet within 6 to 12 months without taking the initiative to slow development. Amodei recommended that all front-line AI commit to providing external assessment teams with "continuous, employee-like access" and that the Anthropic plan be spearheaded, including the provision of office seats, access badges and company laptops; Altman of OpenAI made a similar commitment quickly. In terms of government reviews, the Centre for Artificial Intelligence Standards and Innovation (CAISI) under the Ministry of Commerce was granted advance access to three of the most powerful AI models in the United States on 5 May 2026 to assess potential threats to national security and the public prior to their publication. Previously, CAISI had entered into voluntary agreements with OpenAI and Anthropic, and the new Google, Microsoft and XAI agreements had completed United States top AI list coverage. There is an analysis that compares the urgency of current AI regulation to the state of response in the early stages of the epidemic. In addition, the disclosure of the GPT-6 Astra model used by Junp Trading has enabled the treatment of more vague, difficult and wide-ranging tasks, and AI intelligence has been able to carry out its analytical work autonomously for several days, further exacerbating industry concerns about safety regulations not keeping pace with technological developments. Samuel Marks, a researcher in Anthropic, also noted in his personal analysis that AI developers believed that their technology could lead to the extinction of human beings by 2030 and that the higher the position, the greater the concern of the employees.
For the American people, the impact of these regulatory developments is multidimensional and practical. As one of the most concentrated states in the United States, California has an AI regulatory framework that directly affects science and technology enterprises operating in California, practitioners relying on AI tools and ordinary consumers. The audit and certification systems established by SB 813 and AB 1405 imply that companies providing AI services in California are subject to independent assessment, which may affect the functional design, up-line rhythm and availability of AI products. Chinese in science and technology should focus on the specifics of the voluntary agreement between CAISI and the major AI companies, as well as on the progress of AI Kill Switch legislation at the federal level - - These may change the way businesses use AI tools, thereby affecting the employment market. In the field of health care, the Medicare WiseR AI pre-authorization model has been recorded as an undue denial of compensation, and residents of China who use federal health insurance are required to take care of their rights and interests when making decisions in relation to ASI assisted medical care. It is recommended that attention be paid to the specific rules for the implementation of SB 813 and AB 1405 issued by the California Government Official Network to understand how the AI auditors ' registration form operates and the compliance assessment process; companies and individuals operating or using AI services in California should prepare compliance materials in advance. At the same time, it is worth keeping track of developments in federal legislation, which, if landed, will have a wider impact on AI development and deployment throughout the United States.