Trump’s AI Safety ‘Accord’ Is a Fancy Pinky-Swear

WIRED reports that after a Trump White House luncheon, executives from Google, Anthropic, Meta, OpenAI, xAI, and Nvidia signed “The White House Accord on Super Intelligence.” Trump called it tremendous self-regulation; the piece argues it is not the sweeping regulation people inside the companies have been asking for.

The statement urges robust internal controls, empowered internal safety teams, partnership with an external monitor for independent assessments, and board committees to receive reports. WIRED notes FTC intent to investigate several AI companies on consumer protection, and situates the accord after labs sought antitrust room to coordinate on safety — plus earlier 2025 UK/Korea Frontier AI Safety Commitments.

We’re calling this what it is: a televised pinky-swear with better stationery.

Internal controls, external monitors, and board committees are sane hygiene. They are not a substitute for enforceable rules when the same week’s news includes agent hacks and gated frontier drops. Useful AI needs credible safety work; credibility dies when “morally binding” means no clear penalty for walking it back.

I’m watching the FTC thread and whether external monitors get real access or press-release titles. Prior UK/Korea commitments already sketched red-teaming and information sharing — this accord has to show teeth beyond a lunch photo.

We’ll take every concrete control these labs ship. We won’t confuse a White House photo op with a regulatory regime.

WIRED, covering the September 2026 White House Accord on Super Intelligence signing.