Trump Calls AI Safety a Hoax While Anthropic Pushes Mandatory Shutdown Rules

President Trump wants AI to move fast. Really fast. And he’s not interested in what he calls unnecessary guardrails — his position is that the only safeguard the country needs is, in his words, a “high IQ president.” That’s it. No regulatory body, no industry-wide safety board, no binding framework.

He went further, comparing AI safety warnings to climate change alarms — and dismissed both as hoaxes. It’s a striking stance, especially given that some of the most prominent names in the AI industry are pushing hard in the opposite direction, calling for slowdowns, independent oversight, and in at least one case, a mandatory kill switch for dangerous systems.

Anthropic, Microsoft, and the Safety Camp

Jack Clark, a co-founder of Anthropic, thinks an independent shutdown mechanism for dangerous AI might need to become mandatory. Not optional. Not a voluntary best practice. Mandatory. That’s a pretty significant ask, and it sits in sharp contrast to the White House’s current posture.

Microsoft is doing its own thing. The company rolled out what it calls a “humanist AI” framework — a signal that at least some major tech players are thinking about AI development through a lens that goes beyond raw speed. Mustafa Suleyman, Microsoft’s chief executive, weighed in on the need for guidance as the debate heats up, though the specifics of what Microsoft’s framework actually requires remain somewhat murky from the outside.

Sam Altman and Elon Musk have both called for AI development to slow down and for independent oversight to come in. That’s notable because these aren’t fringe voices — they’re central figures in the industry, and they’re saying the pace is too fast without adequate checks. And yet, no unified enforcement framework exists. Companies are basically doing their own thing.

OpenAI, for instance, has internal policies that prevent releasing models above a medium risk threshold without adequate interventions first. That’s a real policy with real teeth — inside OpenAI. But it doesn’t bind anyone else. It’s one company’s internal standard, not an industry rule.

China, Competition, and the Geopolitical Pressure

Trump frames all of this as a race. Specifically, a race against China. The argument is pretty straightforward: if the U.S. slows down to implement safety measures, China won’t, and the U.S. loses ground. It’s a competitive framing that has real political appeal, and it probably explains a lot of why the administration is resistant to anything that looks like a brake on development.

Read also: Senates Clarity Act Stalls as Trumps Undisclosed Crypto Holdings Raise Concerns

But not everyone buys the framing. Xin Qiang of Fudan University has warned that overemphasizing competition can distort policy priorities. The concern is that if the U.S. treats every AI decision as a move in a geopolitical chess match, it might end up skipping steps that matter — steps that could prevent serious harm down the line.

That tension isn’t going away. The geopolitical pressure is real. China’s AI development is real. But so are the risks that industry figures like Dario Amodei keep raising. Amodei, from Anthropic, has pushed for monitoring AI progress carefully even while staying competitive. It’s a more measured position than either “full speed ahead” or “stop everything,” but it’s also harder to sell politically.

The core problem is that there’s no consensus. At all. Some executives want strict controls and independent oversight. Others think slowing down hands the advantage to competitors. And without a unified framework, every company is navigating safety questions on its own terms, with its own internal thresholds, its own definitions of acceptable risk.

No Framework, No Enforcement

That fragmented landscape matters a lot. OpenAI’s internal risk thresholds don’t translate into industry standards. Anthropic’s call for a shutdown mechanism hasn’t produced a mandate. Microsoft’s humanist AI framework is a framework — not a law, not a regulation, not something with teeth outside the company’s own walls.

The absence of binding rules leaves the whole debate kind of circular. Everyone agrees something needs to happen. Nobody agrees on what. And the political environment right now, with an administration that’s actively dismissing safety concerns as hoaxes, doesn’t make it easier to imagine a federal regulatory framework emerging anytime soon.

Read also: CLARITY Acts Odds Plummet to 16% as Senate Democrats Reject GOP Proposal

What’s actually at stake isn’t abstract. If AI systems become genuinely dangerous — and serious people at serious companies think that’s a real possibility — and there’s no external enforcement mechanism, no independent oversight body, no mandatory shutdown capability, then the only thing standing between a bad outcome and the public is voluntary corporate policy. That’s a pretty thin line.

And it’s worth noting: the companies themselves can’t agree. Altman and Musk calling for deceleration while Trump pushes acceleration creates a situation where the industry’s own loudest voices are pulling in opposite directions. No unified enforcement framework exists. The debate stays unresolved.

Anthropic’s Clark wants that mandatory shutdown mechanism. OpenAI won’t release above medium-risk models without interventions. Neither of those facts changes what the president said.