OpenAI, Google and four others agree to outside AI safety audits in voluntary pact

OpenAI, Google, Meta, Anthropic, Nvidia and Elon Musk’s xAI signed the voluntary White House agreement on September 29. The companies agreed to bring in outside auditors to check their AI safety controls. The agreement carries no penalties if companies fall short. It has no enforcement mechanism and does not require companies to name or publish the auditors. President Donald Trump called the agreement “morally binding” and said the companies must police themselves. Trump said he would set up a 10-member board to oversee AI safety. Trump said he would appoint a new White House official to lead AI policy. The agreement says its measures could eventually be written into law. The one-page document asks companies to monitor their most capable models during training and use for cyberattack, biological and chemical risks. It calls for controls to prevent models from hacking or accessing computer systems in unintended ways. Experimental AI agents have accessed unauthorized systems, and AI has been linked to cryptocurrency security breaches. Attacks have hit software, including bitcoin. xAI is now part of SpaceX. OpenAI President Greg Brockman represented OpenAI at the meeting, alongside Google’s Sundar Pichai, Meta’s Mark Zuckerberg, Anthropic’s Dario Amodei and Nvidia’s Jensen Huang.