California has recently passed a significant AI safety bill, S.B. 1047, that mandates major AI companies to ensure their systems are tested for safety before releasing them to the public. This legislation also empowers the state’s attorney general to sue AI developers for any serious harm caused by their technologies, such as fatalities or property damage.
Historically, California has been a trendsetter in the U.S. when it comes to pioneering legislation. Laws passed in California often influence other states, setting a standard for regulations. As the European Union has already implemented its AI Act, the U.S. has yet to pass federal laws that address high-risk technologies like facial recognition software.
Elon Musk, posting on X, expressed his support for regulating AI, including S.B. 1047, citing the potential risks these technologies pose to public safety.
Supporters of the bill argue that AI has the potential for catastrophic misuse, such as biowarfare or disinformation campaigns that could undermine democratic processes. They believe that AI tools carry substantial risks and that the federal government has not sufficiently addressed these risks through regulation.
SB 1047 is based on the assumption that AI systems in the near future could become extremely powerful and potentially dangerous, necessitating some level of oversight. This fundamental idea is highly debated among AI researchers.
However, there is opposition from major tech companies like Google, Meta, and OpenAI, as well as from Y Combinator and Fei-Fei Li, a notable AI expert advising President Biden. These critics argue that the compliance costs associated with the bill could harm the industry, particularly startups, and deter companies from releasing open-source AI tools due to fear of legal liability.
To address these concerns, California lawmakers have made adjustments to the bill. They removed a proposal to create a new agency specifically for AI safety and adjusted the liability clauses so that companies are held accountable only for actual harms rather than potential ones.
Despite these changes, Meta has maintained its opposition, arguing that the bill could stifle AI development in California, negatively impact business growth and job creation, and contradict the state’s tradition of supporting open-source innovation. They believe that while the bill is well-intentioned, it is not yet ready to be enacted.
Several members of Congress, including Nancy Pelosi, have raised concerns in a letter to Governor Gavin Newsom. They argue that AI is still in its early stages and that it’s premature to determine how these systems could be proven safe.
Opponents of the bill worry that it could stifle innovation and slow the progress of startups. They contend that the legislation would discourage tech giants from sharing their AI software code, a common practice known as open source, which is vital for collaborative development.
Besides S.B. 1047, the California Legislature is working on other AI-related laws. These include requiring large online platforms to remove deepfakes related to elections, establishing a working group to guide schools on safe AI use, and additional measures to combat automated discrimination. Other proposed regulations aim to protect the rights of deceased individuals from having their likenesses used without permission and to safeguard voters from deceptive deepfakes.
References
https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240SB1047
https://www.nytimes.com/2024/08/28/technology/california-ai-safety-bill.html
https://calmatters.org/economy/technology/2024/08/ai-safety-bill-california-legislature/
https://www.vox.com/future-perfect/369628/ai-safety-bill-sb-1047-gavin-newsom-california