On Monday, Governor of California Gavin News signed Transparency in border artificial intelligence into a law requiring artificial intelligence companies to disclose its safety practice, stopping without performing actual safety tests. The law requires that companies with an annual income of at least 500 million US dollars to publish security protocols on its websites and reports on incidents to state authorities, but it lacks stronger teeth. veto Last year after technological companies Very lobbying Against this.
Legislation, SB 53Replaces the previous attempt by Senator Scott Winner to regulate artificial intelligence, known as SB 1047This would require safety tests and “murder switches” for artificial intelligence systems. Instead, the new law asks the company to describe how they include “national standards, international standards and the best practices in the industry” in the development of AI, without indicating that these standards or require independent verification.
“California proved that we can establish the rules for protecting our communities, and also guarantee that the growing AI industry continues to prosper,” the Newsom statement said, although the actual protective measures of the law remain largely voluntary, in addition to the basic reports of reporting.
According to the Government of the state of California, state Houses 32 of the 50 best world companies and more than half of global venture capital for startups and machine learning went to Bay Area last year. Thus, although the recently signed bill is the legislation at the state level, what is happening in the California regulation of artificial intelligence will have a much broader influence both in legislative precedent and due to affecting companies that create AI systems used around the world.
Transparency instead of testing
In cases where the veto, which SB 1047 would make by the obligated security tests and murders for artificial intelligence systems, the new law is focused on disclosure. Companies should report what the staff calls “potential critical security incidents” in the California Emergency Directorate, and provide protection of informants for employees who cause security problems. The law defines a catastrophic risk narrowly as incidents, potentially leading to more than 50 deaths or damage to $ 1 billion. The United States due to the help of weapons, autonomous criminal actions or loss of control. The Prosecutor General can charge civil fines in the amount of up to $ 1 million for violation of non -compliance with these reporting requirements.
