On September 29, California Governor Gavin Newsom (D) vetoed a controversial piece of policy regarding the AI Bill, whose full name The Safe and Secure Innovation for Frontier Artificial Intelligence Models Act, called SB 1047, required companies that develop advanced AI models to comply with a series of safety guidelines, even authorizing potential criminal or civil actions against AI developers. Critics objected to the bill’s focus on catastrophic injuries that would disproportionately affect small developers.
“While SB 1047 has good intentions, it ignores that AI systems are used in high-risk environments, involving critical decision-making or the use of sensitive data. Instead, the bill applies strict standards to even the most basic things—as long as the big plan applies them,” Newsom wrote in his veto letter. “I don’t believe this is the best way to protect the public from the very real threats that technology poses.”
However, the veto does not mean that AI legislation is completely stopped in California or the US at large. AB 2013which needs to be transparent about what data a developer uses to train a productive AI system or service.
Meanwhile, the 2024 legislative session was busy with AI bills, with 45 says (along with Washington DC, Puerto Rico and the US Virgin Islands) introduced AI bills at that time. For example, i The Colorado AI Act requires developers of high-risk AI systems to avoid algorithmic bias. It was the first time it was signed into law in the US, even before the European Artificial Intelligence Act (AI Act). which started working in August.
“I’m sure this legislation will come back in the next legislative session,” said Tatiana Rice, deputy director of the US Legislation at the Future of Privacy Forum, a non-profit, non-partisan organization that is talking about SB 1047. “Next year, maybe the idea of regulating the foundation models of the border will be more mature if there is more moderation at the national or international level.”
At the federal level, any regulatory agency will largely depend on who the next US president is. “Whoever wins the election will determine where AI policy is headed at the federal level, which will have an impact on how states look to fill the gaps,” Rice said.
Ashley Casovan, executive director of the AI Governance Center at the International Association of Privacy Professionals (IAPP), expects industry-specific regulation to reach California and beyond. “What is acceptable is that AI used to diagnose health conditions should be different than AI used to power self-driving cars,” Casovan told the Observer.
Most of the organizations IAPP works with agree that the law should allow for innovation but provide more clarity on oversight. Casovan said this is especially interesting for “those who will be the ones who use and use AI instead of building it,” a camp that many companies fall into.
In terms of enforcement, attorneys general in states with AI bills are playing a different game. Craig Smith, an intellectual property attorney at Lando & Anastasi, LLP, is concerned about how the patchwork of AI legislation without government guidance could play into its implementation and enforcement. “Individual states can place different and potentially inconsistent obligations on the development and use of AI,” Smith told the Observer.
With many tech companies operating abroad, the US can learn from the EU’s AI Law and how it works will continue in the coming months, according to Rice. “That doesn’t mean they have to imitate what the EU is doing,” he said, adding that having some level of standardization in terms of definitions, interpretations and objectives can make a big difference.
