Tech Giants Pledge Independent AI Safety Audits With White House Deal

OpenAI, Google and Meta pledge independent AI safety audits under voluntary White House deal

OpenAI, Google and Meta joined Anthropic, Nvidia and xAI in signing a voluntary White House accord that commits the companies to allow independent audits of their artificial intelligence safety controls.

The agreement, endorsed by U.S. President Donald Trump on Sept. 29, is formally titled the “Joint Commitment on Frontier Responsibilities”. Trump described it as “morally binding,” but the pact is not legally enforceable.

According to the terms described in the document shared by Trump, the signatories agreed to bring in outside auditors to assess whether their AI systems behave as developers intend. The commitments also include safeguards aimed at preventing AI tools from hacking into or accessing technical systems in unintended ways, and addressing high-risk areas such as biological threats.

However, the accord leaves substantial discretion to the companies. It has no enforcement mechanism, does not require participants to disclose audit results publicly, and does not set a deadline for when the checks must be implemented. Companies are also able to choose their own auditors and are expected to address any shortcomings themselves.

Executives representing the signatory firms included Anthropic CEO Dario Amodei, OpenAI President Greg Brockman, Google CEO Sundar Pichai, Meta CEO Mark Zuckerberg, xAI’s Elon Musk and Nvidia CEO Jensen Huang.

The deal matters because it moves AI safety audits from a patchwork of individual company promises toward a shared set of expectations across several of the largest developers and infrastructure providers. It also signals that, at least for now, the White House is emphasizing industry self-policing over new federal rules.

The accord also calls for internal controls to monitor AI models and for an independent board committee to review audit findings, while noting these steps could eventually be written into laws or regulations over time.

In remarks accompanying the agreement, Trump said he was considering a 10-member board to monitor AI safety, though no members were named.