Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
New York State legislators adopted a bill Thursday, which aims to prevent the models of border AIs from Openai, Google and Anthropic from contributing to disaster scenarios, including death or injury of more than 100 people, or more than a billion dollars in damages.
The adoption of the Act on increase represents a victory for the IA security movement, which has Land lost in recent years Like Silicon Valley and Trump administration have prioritized speed and innovation. The defenders of security, including the Nobel Prize winner, Geoffrey Hinton and the AI research pioneer, Yoshua Bengio, defended the Act on increase. If it becomes a law, the bill would establish the first American set of legally mandated transparency standards for the AI frontier laboratories.
The Act on increase has some of the same provisions and objectives as California California Bill of controversial AI, SB 1047, who was ultimately veto. However, the bill of the bill, the New York State senator Andrew Gounardes, told Techcrunch in an interview that he had deliberately designed the law on recovery so that it does not cool innovation between startups or university researchers – a joint critic of SB 1047.
“The window to set up railings quickly shrinks given the speed at which this technology evolves,” said Senator Gounardes. “The people who know [AI] The best known as these risks are incredibly likely […] It’s alarming.
The Act respecting the increase is now directed to the office of the Governor of New York, Kathy Hochul, where the bill could either sign, send it back for amendments, or its orient completely.
If it is recommended, the New York IA security bill would require the largest AI laboratories in the world to publish in -depth security and security reports on their border AI models. The bill also requires that IA laboratories report security incidents, such as the behavior of the AI model or the bad players flying an AI model, if they occur. If technological companies do not comply with these standards, the Act respecting the increase in New York Attorney General to wear civil penalties of up to 30 million dollars.
The Act on the increase aims to justly regulate the largest companies in the world – whether based in California (such as Openai and Google) or China (such as Deepseek and Alibaba). The transparency requirements of the bill apply to companies whose AI models have been trained using more than $ 100 million in computer resources (apparently, more than any AI model available today) and are made available to New York residents.
Although similar to SB 1047, in some respects, the law on recovery has been designed to resolve criticisms of previous AI security bills, according to Nathan Calvin, vice-president of state affairs and general lawyer for encoda, who worked on this bill and SB 1047. In particular, the Law on Relance does not require AI model developers to include a “kill kill Switch ”on their models, nor the Hol the maintenance of the AI traction after the realization of TRA for the AI models for the models for the ai for the ai.
However, the Silicon Valley has considerably rejected the New York IA security bill, the New York State assembly member and the Co-Sponsor of the increase law, Alex Bores, told Techcrunch. The boreholes have qualified the resistance of industry without surprise, but said that the Act respecting the increase would in no way limit the innovation of technological companies.
“The Act on the elevation of the NY is another stupid bill of stupid and stupid AI level which will not harm in the United States at a time when our opponents continue,” said Anjney Midha, general partner Andreessen Horowitz, in a Friday post on x. Andreessen Horowitz, alongside the Incubator Y combinator Startup, was among the fiercest opponents of SB 1047.
Anthropic, the AI laboratory focused on safety that called to federal transparency standards for AI companies Earlier this month, did not reach an official position on the bill, said co-founder Jack Clark Friday post on x. However, Clark expressed certain grievances on the magnitude of the increase law, noting that it could present a risk for “small businesses”.
Asked about anthropic criticisms, the senator of the Gounardes state told Techcrunch that he thought he “misses the brand”, noting that he had designed the bill not to apply to small businesses.
OPENAI, Google and Meta did not respond to the request for comments from Techcrunch.
Another common criticism of the increase law is that the developers of AI models simply do not offer their most advanced AI models in New York State. It was a similar criticism brought against SB 1047, and it is largely what was played in Europe thanks to the difficult regulations of the continent on technology.
Assemblymber Bores told Techcrunch that the regulatory burden on the increase in the increase was relatively light and that, therefore, should not demand technological companies that they cease to exploit their products in New York. Since New York has the third GDP in the United States, withdrawing from the state is not something that most companies would take lightly.
“I do not want to underestimate the political meanness that could occur, but I am very confident that there is no economic reason for them not to put their models available in New York,” said the assembly of Borres.
