Trump’s AI Safety Accord Is Just Another Fancy Pinky-Swear

Silicon Valley's heavy hitters lined up at the White House this week to sign the newly minted Accord on Super Intelligence. Google, OpenAI, Meta, Anthropic, xAI, and Nvidia all sent executives to shake hands, smile for the cameras, and pat themselves on the back for stepping up to the plate. Trump called it an act of tremendous self-regulation. I call it a brilliantly orchestrated distraction.

If you peel back the breathless press coverage, you find an agreement with all the teeth of a wet paper towel. It tells labs they should maybe think about setting up internal controls, or perhaps empower an independent team to make sure their models don't spontaneously decide to hack the global power grid. There are no heavy fines. There are no mandatory shutdowns. There is only a vague suggestion that someone, somewhere, ought to keep an eye on things.

We have seen this exact playbook before. Back in early 2025, similar voluntary promises were inked in the UK and South Korea, filled with the same lofty language about severe threats and information sharing. Nothing fundamentally changed then, and nothing will change now. These labs desperately want an antitrust exemption so they can huddle together under the guise of safety without drawing the ire of regulators. When giant tech conglomerates ask for permission to coordinate behind closed doors, alarm bells should ring.

Real engineering requires rigor, accountability, and hard constraints. It writing a vague pledge to the executive branch is none of those things. While these frontier labs play political theater in Washington, the FTC is quietly lining up investigations into consumer protection issues. That tells you everything you need to know about the actual value of this fancy pinky-swear.