Silicon Valley now faces an unprecedented legal framework to supervise the risks linked to artificial intelligence. The governor of California, Gavin Newsom, promulgated the law “SB 53” of September, presented as a commitment between “stimulus to innovation and protection of public safety.”
Therefore, it imposes on companies that develop the most advanced artificial intelligence models of unprecedented transparency obligations: publication of security protocols, serious incident reports within fifteen days and reinforced protection of warning pitchers.
This text marks an important step, adopted a year after the rejection of a first version considered too restrictive by technological giants. The new legislation occurs, while colossal investment amounts go to AI, transported by key players such as Google, Meta, Openai, Anthrope or Nvidia.
Incidents must be informed
In the heart of the law there is also the obligation to prevent and declare any dangerous behavior of AI models. If a system, for example, knows the effectiveness of its safeguards to facilitate the production of biological or nuclear weapons, the incident must be reported. The promoters of the text ensure that it is a world first, intended to anticipate risks with potentially dramatic consequences.
The project was supported by a group of experts recruited to evaluate the threats linked to AI, including the Fei-Fei LI researcher at Stanford University. This same text also marks the revenge of Democratic Senator Scott Wiener, his initiator: his first attempt had been blocked by the governor in 2024, under the pressure of Silicon Valley entrepreneurs, worried about an innovative escape.
California pioneer
This time, Newsom found a balance, despite past criticism and Trump administration to authorize any regulation of this type in the United States. Finally, the SB 53 Law formalizes and extends commitments that are already voluntarily made by several California giants in the sector. Google, Meta, Openai and Anthrope had pledged to carry out safety tests and establish solid protocols.
Now coded in a legal framework, these efforts become mandatory and give California a pioneering role in the global regulation of artificial intelligence. Before, perhaps, other states? In Europe, where attention also focuses on the supervision of AI, a European regulation (AI) of June 13, 2024 already provides, on paper, obligations for IA systems suppliers for general use (GPAI) and prohibits certain AI systems that undermine fundamental rights. But it could be alone at the beginning.
Source: BFM TV
