news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

According to Bloomberg, OpenAI opposes a California bill that imposes new safety requirements on artificial intelligence companies, joining the ranks of tech leaders and politicians who have recently opposed the controversial legislation.

The San Francisco-based startup said the bill would hurt innovation in the artificial intelligence industry and argued that regulation of the issue should be set by the federal government, not the states, according to a letter to the office of California Sen. Scott Wiener, D-Calif., obtained by Bloomberg News on Wednesday.

The letter also raised concerns that the bill, if passed, could have broad and significant impacts on U.S. competitiveness in artificial intelligence and national security.

SB 1047, introduced by Wiener, seeks to establish what his office calls common-sense safety standards for companies that produce large AI models that exceed certain size and cost thresholds.

The bill, which passed the state Senate in May, would require AI companies to take steps to prevent their models from causing serious harm, such as facilitating the development of biological weapons that could cause mass casualties or economic damage of more than $500 million.

Under the bill, companies will need to ensure that AI systems can be shut down, take reasonable care to ensure that AI models don’t cause disasters, and disclose a compliance statement to the California Attorney General. If companies don’t comply with these requirements, they could be sued and face civil penalties.

The bill has been fiercely opposed by many large tech companies, startups and venture capitalists, who say it is an overreach for a technology still in its infancy and could stifle technological innovation in the state.

Some critics of the bill worry it could drive artificial intelligence companies out of California.

OpenAI expressed similar concerns in a letter to Wiener’s office.

"The AI ​​revolution has only just begun, and California's unique position as a global leader in AI is driving the state's economic vitality," Jason Kwon, chief strategy officer at OpenAI, wrote in the letter. "SB 1047 will threaten this growth, slow the pace of innovation, and cause California's world-class engineers and entrepreneurs to leave the state in search of greater opportunities elsewhere."

According to a person familiar with the company's real estate plans,OpenAI has paused discussions about expanding its San Francisco office due to concerns about the uncertain regulatory environment in California, said the person, who requested anonymity to discuss internal conversations.

In a statement, Weiner defended the proposed legislation and said OpenAI's letter did not criticize any provisions in the bill.

He also said arguments that AI talent is leaving the state make no sense because the law would apply to any company doing business in California, regardless of where its offices are located.

A representative from Weiner's office noted that two prominent national security experts publicly supported the bill.

“SB 1047 is a very reasonable bill that requires large AI labs to do what they have already promised to do, which is to test their large models for catastrophic safety risks,” Weiner said. “SB 1047 is well calibrated to what we know about the foreseeable risks of AI and deserves to be enacted.”

Critics argue that the bill, which requires companies to submit detailed information about their models to the state, will hinder innovation and prevent small open source developers from creating startups for fear of being sued.

Last week, in an effort to address some of the resistance, Wiener amended the proposed legislation to remove criminal liability for tech companies that don’t comply, increase protections for small open-source model developers and eliminate a newly proposed frontier model sector.

Previously, developers could be subject to penalties for perjury if they knowingly submitted false information to the government about their safety plans. OpenAI's competitor Anthropic, known for being more safety-focused than its competitors, had previously said it would support some amendments to the bill. But even after the amendments were introduced, the bill continued to face opposition, including from former House Speaker Nancy Pelosi, who issued a statement calling the bill ill-informed.

A group of Democratic members of Congress also spoke out against the bill.

OpenAI and other tech industry companies have hired lobbyists who have been working on the bill, according to state documents.

OpenAI said in the letter that it had been in contact with Wiener’s office about the bill for several months but ultimately did not support it.

“We must protect America’s AI advantage through a set of federal (not state) policies that provide clarity and certainty to AI labs and developers while protecting public safety,” the letter reads.

OpenAI believes that federal agencies such as the White House Office of Science and Technology Policy, the Department of Commerce, and the National Security Council are better suited to manage key AI risks than California state agencies.

The company said it supports several proposed federal legislation, such as the AI ​​for an Innovative Future Act, which provides congressional support for a new American AI Safety Institute.

“As I have said repeatedly, I agree that ideally Congress would take up this issue,” Weiner said in the statement. “However, Congress has not done so, and we doubt that Congress ever will. Based on OpenAI’s arguments about Congress, California will never pass its data privacy law, and given the lack of congressional action, Californians’ data will have no protections.”

SB 1047 will be voted on in the California State Assembly sometime this month.

If passed, it would eventually move to California Gov. Gavin Newsom’s desk.While Newsom has not indicated whether he would veto the bill, he has publicly spoken out about the need to promote AI innovation in California while mitigating risks.