New York lawmaker revives failed California AI legislation

The New York state legislature is considering the RAISE Act (Responsible AI Safety and Education), a new bill aimed at regulating advanced artificial intelligence systems while avoiding pitfalls that doomed similar legislation in California.

The big picture: The proposed legislation, drafted by Assembly member Alex Bores, targets only the most sophisticated AI models while establishing safety requirements and oversight mechanisms.

• The bill would apply exclusively to advanced AI systems requiring more than 10^26 floating-point operations (FLOPs) and $100 million in training costs
• Safety provisions include mandatory development plans, cybersecurity protections, and third-party audits
• The New York attorney general would have authority to impose fines or halt development of unsafe systems

Key provisions and protections: The RAISE Act incorporates several measures designed to prevent potential dangers from advanced AI systems while protecting those who identify risks.

• Companies must implement comprehensive safety plans for model development and deployment
• Whistleblower protections are included for employees who report safety concerns
• The bill specifically defines “critical harm” scenarios, such as weapons creation or mass casualty events
• Required risk assessments and third-party audits would provide additional oversight

Learning from past attempts: Drawing lessons from California’s failed SB 1047 bill, the RAISE Act takes a more focused approach to regulation.

• The legislation avoids creating new government agencies or bureaucracy
• No requirements for public cloud computing infrastructure
• Eliminates controversial “kill switch” provisions
• Shifts focus from “advanced persistent threats” to more clearly defined critical harms

Points of contention: Some stakeholders have raised concerns about the bill’s scope and priorities.

• Critics argue the focus on catastrophic risks overlooks immediate AI-related issues like algorithmic bias and workforce displacement
• The tech industry may oppose new regulatory requirements
• The high threshold for affected systems means many AI applications would fall outside the bill’s scope

State-level innovation in AI policy: In the absence of federal AI regulations, New York’s proposal represents a significant attempt at establishing guardrails for advanced AI development through state legislation.

The proposal faces an uncertain path forward as stakeholders debate whether this targeted approach to regulating the most powerful AI systems strikes the right balance between innovation and safety. While the bill addresses some criticisms of previous legislation, questions remain about its effectiveness in addressing the full spectrum of AI-related risks.