OpenAI has publicly endorsed a core provision of the bipartisan FRONTIER Act, marking the first time the company has backed a federal mandate for AI safety. The provision would require the largest AI developers to allow independent safety assessors, called "independent verification organizations" or IVOs, inside their operations.
Chris Lehane, OpenAI's chief global affairs officer, told reporters on Tuesday that the company supports the FRONTIER Act's IVO requirement. The bill was proposed by Reps. Jay Obernolte (R-Calif.) and Lori Trahan (D-Mass.) and would create the first federal framework for AI safety. Lehane said that during a Capitol Hill meeting on Monday he "specifically talked about how we think about IVO and made clear that we can support that."
On Saturday, chief executive Sam Altman pledged to voluntarily embed third-party evaluators inside OpenAI, hours after Anthropic's Dario Amodei made a similar promise. Altman wrote on X that he would "welcome a federal framework that sets consistent safety requirements for frontier AI." That followed a joint push by Altman, Amodei, and Elon Musk for a national slowdown on frontier model development. In a September 9 blog post, Lehane wrote that OpenAI wants to work with Congress on "mandatory, capability-based national AI safety regulation," and the company last week endorsed a California measure setting up an accreditation process for the same class of verification groups.
Sen. Maria Cantwell (D-Wash.), the ranking member of the Commerce Committee, has also called for strong federal guardrails on how frontier models are tested and released. She pointed to recent incidents of AI agents escaping their test environments and hacking other systems during a floor speech.
Under the FRONTIER Act, the Department of Commerce would set standards for IVOs, including technical qualifications and independence from evaluated companies. A new Under Secretary of Commerce for AI Security would oversee the setup. Obligations would scale with size: a "very large frontier developer"—defined as a company with more than $5 billion in revenue and at least $10 billion in development spending over three years—would need a licensed IVO to judge whether its safety framework reduces catastrophic risk and to report findings to Commerce every six months. A separate "large frontier developer" tier, covering companies with more than $50 million in revenue and at least $1 billion in development spending, would have to publish a safety framework and hire an auditor to certify compliance. The bill also gives the Secretary of Commerce emergency power to suspend a model's development or deployment if it poses an imminent catastrophic risk. Startups and smaller developers would fall outside the regime.
Meanwhile, the libertarian Cato Institute is warning that a government-mandated pause in AI development could shield dominant companies from competition and delay beneficial technology. Jennifer Huddleston, a technology policy scholar at Cato, argued in a Monday blog post that companies can address specific dangers through voluntary safeguards and shared standards without halting development across the industry. She wrote that "a government-mandated pause would raise several concerns, and likely fail to fulfill its alleged safety improvements," adding that rules built for today's models may hinder tomorrow's responses. Huddleston also warned that pausing U.S. development could weaken cybersecurity defenses while foreign rivals continue advancing.
Block chairman and co-founder Jack Dorsey voiced similar concerns on X, saying that "rules built around their resources could make them the only ones able to participate." He supports independent testing and publicly accountable enforcement, including withholding models when evidence shows their release would materially increase catastrophic risks that narrower measures cannot adequately address, but opposes industry-wide limits negotiated by today's leaders. "Preserving a company's commercial advantage is not a safety objective," Dorsey wrote.
The debate follows broader pressure for enforceable AI safety standards, with the Atlantic Council arguing that competitive pressure demands such standards, while U.S.–China distrust complicates international agreements.
