California, headquarters of the leading American technology companies such as Google, Meta, Openai and Anthropic, applies a new new law on Monday regulation of artificial intelligence (Ai). This law comes a year after the rejection of an earlier proposal that had been particularly restricted to innovation.

Governor Gavin Newsom has announced that he has signed the SB 53 law, stressing that he is achieving a balance “that helps the California state to encourage innovation while protecting public security at the same time.”

The new law is implemented at a time when billions of dollars are heading to TN at Silicon Valley, while increasing concerns about the potential dangers of the most sophisticated models.

It provides for strict transparency obligations for companies developing advanced artificial intelligence systems. Businesses are called upon to publicize their security protocols, report serious incidents within 15 days and protect the witnesses of public interest.

In addition, the law requires reporting of dangerous or misleading behaviors of TN during testing. For example, if a model conceals evidence of controls aimed at preventing its involvement in the construction of biological or nuclear weapons, the incident must be disclosed when it significantly increases the risk of damage.

“Reports of TN’s state-of-the-art businesses reveal alarming advances in all categories of threats,” the expert team set up under the auspices of the Governor said in June, with the participation of Fey Fai Lee from Stanford University, known.

Senator Scott Wiener, author of the law, had seen his original plan rejected by the commander last year. The first version had sparked reactions to Silicon Valley, with executives fearing that it would discourage innovation at the beginning of the “TN revolution”.

The new law entered into force after the Trump government’s failure to prevent US states from regulating the field of artificial intelligence, arguing that this would harm competitiveness over China.

According to Senator Wiener, the new framework ensures the publication of the security protocols, while the corresponding European legislation restricts their transmission only to the authorities.

Before passing the law, many California companies, including Meta, Google, Openai and Anthropic, had already volunteered for security tests and adopting strict protocols. The law is now instituting and reinforcing these commitments.