Trump, Johnson to meet AI executives at White House

Strip away the theater and what's actually happening here is a negotiation over who writes the rules for a technology moving faster than any regulatory body can currently track. That's a legitimate governance problem, and it deserves better than what's on offer from either side of this meeting.

Start with the framing. Trump calling existential-risk warnings a 'hoax' is not a risk assessment, it's a dismissal. Nobody has definitive proof that advanced AI poses civilizational danger, but nobody has definitive proof it doesn't, either — the honest position is uncertainty, not a punchline. Dismissing the concern outright discourages exactly the kind of independent, technical evaluation that would let policymakers act on evidence rather than vibes.

Johnson's preferred fix — let companies regulate themselves — has a track record, and it isn't reassuring. Self-regulation has already been tried, in social media, in data privacy, in content moderation, and the results were inconsistent enforcement, opaque standards, and public trust that kept eroding until Congress and regulators stepped in anyway, usually too late to prevent the damage. There's no obvious reason AI safety pledges from the same companies would behave differently, especially when the people making the pledges are also the people competing to ship product fastest.

The China argument is real but should not function as an automatic trump card. Competitive pressure is a genuine constraint on unilateral slowdown — that's a fair point. But it's also the standard justification companies and governments reach for whenever anyone proposes a safeguard that might cost time or money. The claim deserves scrutiny, not automatic deference: what specific safeguards would meaningfully hand China an advantage, versus which ones are just inconvenient for U.S. firms? That's an empirical question this meeting isn't equipped to answer, because it's a room full of executives with a direct financial stake in the answer being 'none.'

Notice also who's in the room and who isn't. This is CEOs and elected officials, not technical standards bodies, not independent safety auditors, not the agencies that would actually have to enforce anything. A dinner and a photo-op is not a regulatory framework. If the outcome is a voluntary commitment with no enforcement mechanism, it will function as reputational cover, not governance.

The more defensible near-term approach — and one this meeting doesn't seem oriented toward — is targeting concrete, measurable harms: fraud, labor disruption, security vulnerabilities in critical systems, deceptive content. Those are addressable with disclosure requirements, liability standards, and testing regimes enforced by agencies with actual authority, not press-conference pledges from the companies being regulated. The informational website is a low-cost, sensible step; a serious AI policy is not.