Source: Wall Street See

This week, the controversial AI bill in California, USA, will enter the final voting and signing stage. Once implemented, mainstream AI large models may face stricter regulation.

The bill, known as the 'Frontier Artificial Intelligence Model Security Innovation Bill' (SB-1047), aims to establish clear security standards for AI large models with training costs exceeding $0.1 billion or reaching a certain computing power. It requires developers to conduct security testing before deployment, simulate hacker attacks, install cybersecurity safeguards, and provide protection for whistleblowers. The bill will allow California law enforcement to hold developers of large models accountable for causing 'serious harm', such as causing multiple casualties or losses exceeding $0.5 billion.

Since its proposal in February this year, the bill has continued to trigger controversy. Many AI researchers, entrepreneurs, and tech giants believe that many provisions of the bill are unreasonable and will stifle innovation. Supporters, on the other hand, believe that AI should be constrained and require a set of legal standards for regulation.

The bill has also received support from Musk, who owns an AI company, stating on social media that the California AI bill should be passed. OpenAI, which had previously expressed opposition, has changed its stance and expressed support for certain provisions of the bill. Noted scholar and Stanford University computer science professor Fei-Fei Li, who opposes the bill, stated that it does not address the problem it intends to solve and will limit innovation.

Reports indicate that the bill has previously passed in the state senate and is expected to be voted on in the state assembly before this weekend. Due to multiple revisions, the bill will return to the legislature for a final vote. If it is passed, it will be sent to California Governor Gavin Newsom for signing.

The California AI bill has sparked division in Silicon Valley.

California State Senator and co-author of the bill, Scott Wiener, accused some opponents of using 'spreading panic' to block the passage of the state legislation.

Wiener stated: 'Some opponents have made many dramatic and incorrect statements. Large laboratories have repeatedly committed to conducting security assessments of their (AI) large models, which is exactly what the bill requires. The purpose of the bill is to transform voluntary commitments into mandatory requirements.'

Wiener held an online press conference this Wednesday with supporters of the bill, including renowned AI 'father' Yoshua Bengio, AI Security Center Director Dan Hendrycks, and CEO Thibault Duchemin of AI startup Ava.

On August 27, Musk expressed support for the California AI bill on social media. He said, 'This is a difficult decision that will leave some people dissatisfied, but all things considered, I support California passing SB 1047 AI safety bill. For over 20 years, I have been an advocate for AI regulation, just like we regulate any product/technology that carries potential risks to the public.'

Despite OpenAI's previous explicit opposition to the California AI bill, the company's Chief Strategy Officer Jason Kwon expressed support for certain provisions of the bill in a letter to Wiener. However, Kwon still believes that AI regulation should be determined by the federal government of the USA, rather than pieced together by state laws.

Recently, OpenAI publicly expressed opposition to the bill, with the company's Chief Strategy Officer Jason Kwon stating that the bill threatens 'California's unique position as a global leader in artificial intelligence.'

After proposing a series of modifications to the bill, OpenAI's competitor Anthropic cautiously acknowledged the bill. The company's CEO stated that the bill currently 'has more pros than cons.'

At the same time, the bill also faced strong criticism from scientists like Fei-Fei Li and leading AI companies. Fei-Fei Li, a computer science professor at Stanford University, expressed that the SB 1047 bill does not address the issues it aims to solve, and would severely harm the AI academic community, open-source community, etc., as developers would bear excessive responsibility, restricting their innovation.

Google, Meta, and Andreessen Horowitz, as well as other major technology companies, also believe that this bill poses a threat to innovation and research. Some people are concerned that this could put California at a disadvantage and may even lead to the loss of AI companies located in Silicon Valley.

Editor / jayden