OpenAI Urges U.S. Congress to Replace Voluntary AI Safety Pledges With Mandatory Law

OpenAI is urging the U.S. Congress to enact mandatory, capability-based national AI safety regulations to govern the development of frontier models. The move signals a decisive shift away from the industry's previous reliance on voluntary safety commitments.

The proposed federal framework would move beyond promises, establishing compulsory testing standards and independent third-party audits for developers. According to the company, the regime must also include enhanced cybersecurity defenses and strict incident-reporting protocols to ensure transparency when safety failures occur.

The Shift to Mandatory Governance

For years, the AI industry has operated under a system of voluntary pledges. However, as models evolve toward "agentic" behavior—where AI can independently take actions within external systems—the risks associated with autonomous operation have grown. This technical evolution has created a fragmented regulatory environment where individual states are passing laws in the absence of a unified federal standard.

In the interim, OpenAI is backing four specific state-level bills in California to establish a baseline for safety. These include SB 813, which focuses on infrastructure for independent safety assessments, and AB 1405, which sets standards for AI auditors. The company is also supporting SB 1119 to protect young people and AB 1864, which aims to create safeguards against AI-enabled biological threats.

Why the Pivot Matters

This policy reversal from one of the world's leading AI developers marks a critical turning point in global AI governance. By admitting that voluntary measures are insufficient, OpenAI is effectively signaling that the pace of AI capability is outstripping the ability of private companies to self-regulate.

Industry analysts suggest this push could pave the way for a formal U.S. federal AI regulatory agency or a stringent licensing regime for "frontier" models. Such a shift would fundamentally change how AI companies operate, moving them from a self-policed ecosystem to one overseen by government-mandated compliance and legal accountability.

The Path Forward

The urgency of the request is tied to the potential for AI to accelerate its own development. "The prospect of AI-accelerated AI development demands more than voluntary commitments," said Chris Lehane, OpenAI Chief Global Affairs Officer. "The United States needs mandatory, capability-based national regulation that can evolve as the technology does."

What remains to be seen is whether Congress will act to create a centralized federal authority or if the U.S. will continue to rely on a patchwork of state laws. As competitors like Anthropic also navigate potential public offerings and increasing scrutiny, the industry's appetite for a predictable, mandatory legal framework is likely to grow.