The Governor of California, Gavin Newsom, has overruled a contentious proposal aiming to govern artificial intelligence (AI). He expressed worry that the legislation could hinder progress, a stance he reached amid intense lobbying by technology firms.
Newsom, a member of the Democratic party, delayed until the last moment to declare this decision, subsequent to the passage of the legislation via the state legislature at the tail end of August.
This legislation would have obliged those responsible for designing superior AI models to comply with rigorous rules. This included the incorporation of a termination function to mitigate the risk of catastrophic damage.
Prominent AI organisations, such as Google, OpenAI, and Meta, all objected to the legislation, heavily campaigning against it. They claimed that this untimely legislation might hinder AI development and compromise California’s frontier position in this technological field.
Meanwhile, Anthropic, backed by Amazon, and Elon Musk, the owner of start-up xAI, endorsed the legislation.
When defending his veto of the Safe and Secure Innovation for Frontier Artificial Intelligence Systems Act, also known as SB 1047, Newsom drafted a letter to the state senate. In this letter, he suggested that such a framework could “inhibit the very innovation which drives advancement in the interest of societal welfare”. He highlighted the fact that 32 of the planet’s leading AI companies are located within California.
He voiced concerns over the unreasonable emphasis on model size by the legislation – it enforced safety testing and other safety measures for models that exceeded $100 million in development cost. Newsom proposed that this could misleadingly give the public a sense of being able to manage this rapidly-evolving technology, ignorant of how “smaller, specialised models might arise as just as hazardous, if not more”.
Scott Wiener, the senator responsible for proposing the bill, viewed Newsom’s veto as a major reversal for everyone advocating the supervision of sizable corporations making key decisions that could impact public welfare and the planet’s future.
He also argued that companies intending to develop a highly potent technology escape any binding limitations from US policymakers, especially considering the ongoing inaction of Congress towards the effective regulation of the tech industry.
Nonetheless, Newsom claimed the bill did not consider whether an AI system “operates in high-risk settings, involves critical decision-making or the usage of sensitive data”. He added, “the bill imposes stringent standards on even rudimentary functions – as long as a sizeable system applies it. I am not convinced that this is an ideal approach to shield the public from the tangible threats posed by the technology.”
Over the last month, bills have been passed by Mr Newsom concerning the implementation and administration of generative AI technology. This type of technology, which generates text and imagery, includes matters such as deepfakes, AI watermarking, and misinformation. He detailed that experts in the technology field were working collaboratively with the state to devise effective protective measures for the application of generative AI, supported by empirical and scientific evidence.
However, the governor’s veto was dubbed as “misguided, thoughtless and not in harmony with the citizens he is supposed to govern” by the Artificial Intelligence Policy Institute, a renowned think-tank. Daniel Colson, the executive director, implied that Mr Newsom missed an opportunity to exhibit leadership in the regulatory governance of AI development, a road he has taken with other sectors. Instead, he has chosen a hands-off approach, risking uncontrolled deviation of AI development.
Colson urged for Mr Newsom and fellow lawmakers to reconvene in Sacramento during the following session to reach a consensus on a collection of provisions that will place rational protective measures on the growth of AI. – Copyright The Financial Times Limited 2024.