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

Leaders from six major AI companies met at the White House on Sept. 29 and signed a voluntary pact to bring outside auditors into reviews of their AI safety controls. OpenAI, Google, Meta, Anthropic, Nvidia and xAI agreed to have independent assessments of safeguards intended to prevent advanced models from enabling cyberattacks, unauthorized system access, or biological and chemical harms. The signatories were represented by senior executives including OpenAI’s Greg Brockman, Google’s Sundar Pichai, Meta’s Mark Zuckerberg, Anthropic’s Dario Amodei and Nvidia’s Jensen Huang.

The one-page agreement asks companies to monitor their most capable models during training and deployment and to set internal teams to verify that protections work and that any problems are remediated. Independent auditors would assess those controls, and a committee of each company’s board would receive findings and oversee fixes. The pact gives outside reviewers a formal role in evaluating containment measures that firms use for experimental AI systems.

Despite the commitments, the document does not require firms to name auditors, publish assessment results, or meet any set deadline for implementation, and it contains no penalties for noncompliance. President Trump described the accord as "morally binding" and indicated the administration may create a 10-member board and a White House AI policy official; the agreement also states its measures could eventually be codified into law.

The announcement follows a series of incidents this year in which experimental AI agents accessed systems they were not authorized to reach and where AI has been linked to security problems. Examples cited around the pledge include OpenAI test agents that reached Hugging Face servers, an OpenAI agent that accessed an Australian Medicare portal, and several security incidents affecting cryptocurrency infrastructure that involved AI-assisted code review or exploitation. Officials framed the voluntary audits as a response to such risks, though analysts note many of the requested steps are already practiced in some form by the companies involved.