New York Lawmakers Approve RAISE Act to Enhance AI Safety Standards
On Thursday, New York state legislators passed a significant piece of legislation known as the RAISE Act, aimed at ensuring that advanced AI models from organizations such as OpenAI, Google, and Anthropic do not contribute to catastrophic outcomes. These outcomes could potentially involve the loss of over 100 lives or result in damages exceeding $1 billion.
A Step Forward for AI Safety
The enactment of the RAISE Act is a notable victory for proponents of AI safety, a movement that has faced challenges in recent years as the emphasis on rapid innovation has surged, especially in Silicon Valley and during the Trump administration. Noted figures in the field, including Nobel Prize winner Geoffrey Hinton and AI pioneer Yoshua Bengio, have been strong advocates for this legislation. If ratified, the RAISE Act would introduce the first legally mandated transparency standards for leading AI laboratories in the United States.
Key Provisions of the RAISE Act
This new legislation shares some similarities with California’s contentious AI safety bill, SB 1047, which faced a veto. However, New York state Senator Andrew Gounardes, a co-sponsor of the RAISE Act, noted in an interview with TechCrunch that he consciously crafted the bill to avoid hampering innovation among startups and academic institutions, a concern often directed at SB 1047.
“The opportunity to implement regulatory safeguards is diminishing rapidly as technology continues to advance,” warned Senator Gounardes. “Experts in the AI field indicate that the risks are significantly probable, which is alarming.”
The RAISE Act is now poised to be reviewed by New York Governor Kathy Hochul, who has the options to endorse the bill, suggest amendments, or veto it entirely.
Transparency Requirements and Potential Penalties
Should the RAISE Act become law, it would necessitate that major AI organizations publish comprehensive safety and risk assessment reports regarding their advanced models. Furthermore, AI labs would be obligated to report any safety incidents, including unusual AI behaviors or instances of theft involving their models. Non-compliance could lead to civil fines of up to $30 million, enforceable by New York’s attorney general.
The legislation specifically targets some of the world’s most prominent companies, regardless of their geographical base—be it California’s tech giants or China’s emerging corporations. The transparency stipulations apply to companies whose AI models consume over $100 million in computing resources, a threshold likely met by the most advanced AI systems currently in circulation.
Addressing Criticisms and Industry Reactions
The RAISE Act is designed to incorporate feedback and criticisms directed at earlier proposals, according to Nathan Calvin, vice president of State Affairs and general counsel at Encode, who contributed to both this bill and SB 1047. Notably, it does not mandate a "kill switch" feature in AI models, nor does it impose liability for severe damages on companies that post-train advanced AI systems.
Despite the design considerations, there has been significant opposition from Silicon Valley. State Assemblymember and co-sponsor Alex Bores remarked on the unsurprising nature of this pushback but emphasized that the RAISE Act would not stifle technological advancements.
Concerns have been voiced regarding the potential impact of the RAISE Act on innovation. Anjney Midha, a general partner at Andreessen Horowitz, criticized the bill, suggesting it could stymie progress at a time when global competitors are advancing rapidly.
Broader Implications and Future Outlook
Anthropic AI, a company focused on safety measures, has withheld formal commentary on the RAISE Act. However, co-founder Jack Clark expressed concerns that its broad scope could adversely affect smaller enterprises in the industry.
In response to such criticisms, Senator Gounardes emphasized that the bill is not intended to burden smaller companies, highlighting that its regulations are tailored for larger entities.
An ongoing debate revolves around whether advanced AI developers would withdraw their most sophisticated models from the New York market, a concern previously raised against SB 1047. Assemblymember Bores addressed this skepticism by asserting that the regulatory framework of the RAISE Act is manageable, thereby encouraging operational continuity for tech firms within New York. He noted that with New York holding a substantial share of the U.S. economy, exiting the state would be a significant consideration for most companies.
“I recognize the potential for political maneuvering, but I am highly confident that there are no economic deterrents preventing AI companies from offering their models in New York,” concluded Assemblymember Bores.
As the bill awaits the governor’s decision, its passage could pave the way for new standards in AI safety and transparency, impacting both industry practices and regulatory approaches nationwide.
