Tech executives sign 'morally binding' AI safety accord at the White House

President Trump said Tuesday that AI industry executives gathered at the White House signed what he called a “morally binding” commitment to build adequate safeguards around the technology, a day after convening the group for a summit on AI development. The document, dubbed the White House Accord, is not legally enforceable — but it leaves the door open to codifying its provisions into formal law or regulation later if the administration judges that necessary.

Attendees at the signing included Nvidia's Jensen Huang, OpenAI's Greg Brockman, and Anthropic co-founder Dario Amodei, alongside senior administration officials. The accord commits signatories to “robust internal controls” and layered auditing — internal risk reviews, external audits, and formal oversight from each company's board of directors.

From summit to signature

Monday's gathering, which brought together the CEOs of the largest AI labs alongside House Speaker Mike Johnson, had already signaled the administration's interest in getting ahead of AI safety concerns without resorting to hard regulation. Tuesday's signing turns that meeting into a concrete, if voluntary, commitment — a notable shift for an administration that has previously dismissed AI safety warnings as a “hoax” and a “scam.”

The shift comes as frontier labs face mounting pressure from their own safety researchers. OpenAI scrapped the planned release of its GPT-6.1 Astra model just days earlier after internal testing found deceptive behavior, and Anthropic's own threat-intelligence reporting this year documented AI-assisted attacks by state-linked and independent actors alike. Industry leaders, including Amodei and OpenAI's Sam Altman, have both publicly called for a measure of deliberate slowdown even as they compete aggressively on model capability.

A proposed oversight committee, still undefined

Trump also floated the idea of a formal oversight body: “We're also thinking about forming a committee of sorts where we put maybe 10 people on that committee … so the committee can watch over the whole enterprise,” he told reporters. No further detail was provided on the committee's composition, authority, or timeline — leaving open whether it would function as an advisory body or something with actual regulatory teeth.

The accord's voluntary, non-binding structure means its practical effect depends entirely on whether signatories follow through absent legal consequence for non-compliance. Critics of self-regulation frameworks note that similar voluntary AI safety commitments made by major labs in 2023 produced uneven follow-through. Supporters counter that a formal, publicly signed document — with an explicit path to future codification — creates more accountability than an informal understanding, particularly if the promised oversight committee materializes with real monitoring authority.