AI giants accept outside audits in White House pact

OpenAI, Google, Meta and three other AI companies agreed to let independent auditors assess their safety controls under a voluntary White House pact, CoinDesk’s Shaurya Malwa reported. The deal carries no penalties, does not require disclosure of auditors and sets no implementation deadline.

Anthropic, Nvidia and xAI also signed. The accord asks companies to monitor powerful models for cyberattack and biological or chemical threat capabilities, test safeguards internally, and give audit findings to board committees overseeing fixes.

For decentralized AI, the governance question is whether safety controls can become independently verifiable across open, distributed systems. This pact adds outside review, but company-selected auditors, voluntary compliance and no disclosure requirement leave accountability concentrated with the firms themselves.