OpenAI supports a legislative initiative to subject artificial intelligence models to mandatory external audits

OpenAI has expressed support for key aspects of a bipartisan proposal in Congress that would require major artificial intelligence developers to subject their language models to external and independent audits.

Chris Lehane, the company's director of global affairs, outlined the position, which a company spokesperson confirmed to CBS News following a roundtable discussion with lawmakers in Washington, D.C.

This endorsement marks the first time the California-based company has explicitly supported a federal mandate requiring third-party evaluators to review developers' work.

The measure is included in the FRONTIER Act, a bill jointly sponsored by Republican Representative Jay Obernolte and Democrat Lori Trahan.

External oversight and incident reporting

The legislative text seeks to have independent bodies verify whether companies' safeguards are sufficiently rigorous to mitigate what the legislation defines as acceptable levels of catastrophic risk.

Furthermore, the legislation proposes requiring creators of this technology to publish detailed reports on their models' performance and report any significant security incidents.

An OpenAI spokesperson told CBS News that while the company supports the principle of third-party audits, it has not endorsed the legislative proposal in full.

Toward a national regulatory framework

Lehane's position on Capitol Hill complements a previous public statement in which he called for regulatory action to guide the sector's development.

OpenAI has proposed that Congress establish mandatory, binding national security requirements and that industry leaders collaborate to set industry standards.

"The stakes are enormous. Advanced AI could accelerate the development of new medicines, strengthen critical infrastructure, expand economic opportunities, and help solve scientific problems that have defied generations of human effort," Lehane wrote, setting out the corporate position.

"But the capabilities that make these models more useful also carry risks, and they will not remain confined to a few cutting-edge laboratories."

For their part, industry analysts point out that the debate over the control of the technology requires government intervention to establish clear rules of the game.

Joseph Miller, an AI safety researcher and founder of the group PauseAI UK, told CBS News that it is unlikely that companies will voluntarily adopt strict safeguards without the involvement of public authorities: "We need the government to take the lead on this, because they probably won't do it without government intervention."

This week, President Trump spoke out against slowing down AI development through bureaucratic hurdles imposed by the federal government.