A ‘Morally Binding’ White House Accord on AI Safety

AI industry leaders were invited to the White House, where they signed the White House Accord on [Artificial] Intelligence, a Joint Commitment on [AI] Responsibilities. According to Andrew Curran, Speaker Johnson said after the meeting that everyone signed the accord, which includes voluntary commitments, robust internal control, and layers of internal and external review, with the White House and Congress continuing to guide the industry's development in a safe manner. Asked whether the accord was binding, President Trump said, "I think it's morally binding, yeah."

Signatories include Sundar Pichai, Dario Amodei, Mark Zuckerberg, Elon Musk, Jensen Huang and Greg Brockman; Sam Altman was at Dev Day. Meta, xAI and Nvidia, which had opposed the planned SAFA standards group led by Google, OpenAI and Anthropic, signed. Microsoft and Amazon were present and did not visibly sign; it is unclear whether they chose not to sign or were not included.

The accord's text states each company should implement four layers of controls and audits: robust internal controls monitoring model capabilities and alignment during training and deployment around areas like cybersecurity, biosecurity and chemical threats, and ensuring models do not hack or access technical systems in unintended ways; an empowered internal team; an independent external auditor or evaluator; and an independent board committee overseeing reports and remediation. The participating companies will meet regularly to establish standards and best practices, and the text says it may over time make sense to codify these steps into laws or regulations.

Separately, the FTC has issued a CID confirming an industry-wide probe opened this summer, with information demands to at least OpenAI, Anthropic and METR, with at least some relation to the HuggingFace Incident. Trump also said a document would be signed renaming Artificial Intelligence to [Artificial Intelligence], which he claims lab leaders signed off on. Trump said there would be "tremendous self-regulation," and Speaker Johnson said on Squawk Box that "a little oversight" through transparency and external auditing on AI would be appropriate. The account also states that an attempt at a moratorium may come during the lame duck session, with preemption talks focused on state laws on catastrophic risk.