California lawmaker behind SB 1047 reignites push for mandated AI safety reports

California State Senator Scott Wiener, Wednesday was introduced New adjustments The most recent bill will require SB 53 The world’s largest AI companies to publish security and security protocols and the security of security occurs.

If he entered the law, it will be the first state to apply meaningful transparency requirements to the leading AI developers, including California, AI developers, Google, Anthropic and Xai.

Senator Wiener’s Previous AI Bill, SB 1047, AI Model Developers include similar demands to publish security reports. However, Silicon Valley fought violently against this bill and it was As a result, vetoed by Governor Gavin Newsom. A group of California’s governors AS AI leaders – a group of leading researchers and co-founder Fei Fei LII, Fei Fei Li and the state aim for AI security efforts.

California’s AI policy group recently published their Last recommendationsTo create a “healthy and transparent evidence environment”, he asked for “demands to publish information in the industry”. Senator Wiener office made a press release that SB 53 was very impressed by this report.

“The bill continues, and I look forward to working with all stakeholders in the coming weeks to make the most scientific and just law in order to become the most scientific and just law,” Senator Wiener said.

SB 53 Governor’s news aims to strike a balance to create meaningful transparent requirements for the largest AI developers for the largest AI developers for the largest AI developers from preventing the rapid growth of the AI ​​industry.

“These are concerns that my organization and others are talking for a while,” he said. “Companies feel like a reasonable, reasonable step to take these risks to the public and the government to take these risks.”

The bill also creates a “critical risk” to contribute to the deaths or injury of more than 100 people in the bill, or injure more than 100 people in the bill in the draft of AI laboratories.

In addition, the bill is aimed at creating a public cloud computing group to support the starting and researchers developing large-scale AI.

Unlike SB 1047, the Senator Wiener’s new bill is responsible for the losses of AI models. SB 53 was not to create a burden on the starting and researchers who use the EU’s leading AI developers from the leading AI developers and use open source models.

With the new amendments, in SB 53 to consumer protection for agreeing and approval to the California State Parliament Assembly Committee. If the bill is passed there, the bill must undergo several other legislative bodies before the governor’s table reaches the table.

On the other side of the United States, New York Governor Kathy Hochul Now Considering similar AI security bill The increase in the increase in the demanding to publish the security and security reports of large AI developers.

The state is in short as the EU law and the fate of the 53rd SB 53 as a destiny Federal MPs were considered a 10-year AI moratorium State EU regulation – Attempt to restrict Patchwork “Patchwork” of the AI ​​laws, where the companies should depart. But this suggestion Failed in the 99-1 Senate vote in July.

“The provision of the EU is reliably developed, it should be controversial – it must be foundation,” he said. “Congress requires transparency and accountability of companies to establish border models and non-guard models. However, women are a serious federal movement, the states are a thoughtful, well-established example of the state management.”

To this point, MPs could not reach state-owned state companies on state companies with transparency requirements. Anthropically widely confirmed AI needs to increase transparencyand even expressed Modest optimism about recommendations From the AI ​​policy group of California. But companies such as Openai, Google and Meta have become even more sustainable for these efforts.

Leading AI model developers usually publish security reports for AI models, but have been less consistent in recent months. Google, for example, decided Not to release the security report for the most advanced AI model, which is so far, Gemini 2.5 Pro, after a few months. Openai also decided Not to release a security report for the GPT-4.1 model. Then, it claimed that this was a third-party investigation are less adapted than previous AI models.

SB 53 represents a toned version of previous AI security documents, but it can force AI companies to broadcast more information from today. So far, Senator Wiener will closely monitor these borders.