Trump, Google, Anthropic, Meta sign AI superintelligence accord

The meeting and the signing were closely held: the main documentation of who attended comes from an image posted on X by AMD CEO Lisa Su and a video from aide Margo Martin, which places Anthropic's Dario Amodei at the left end of the table. That matters because the accord is a commitment by named people, not a piece of legislation, so the record of who was in the room is the record of who is bound.

The four points move in one direction: from internal controls outward to external validation. Companies are asked to monitor their own models during development and deployment, give internal oversight teams real authority, bring in outside independent auditors, and then route those findings to a new independent board committee. OpenAI is represented by President Greg Brockman rather than a CEO, and SpaceXAI by Elon Musk, whose other ventures have their own AI work.

The stakes hinge on the last two points. Internal controls and internal teams are self-reported by design, so the question a reader should hold is whether the external audits and the board committees are given enough independence to flag a failure and see it corrected. Nothing in the accord as described establishes an enforcement body, so its practical weight is likely to rest on how seriously each signatory treats those reviews.