Trump Calls Amodei a “Perfect Little Angel” — Rejects the AI Pause
After Amodei, Altman, and Musk asked to pace the frontier with embedded evaluators, Trump says the scare is overcooked and America already has the criminal and regulatory power it needs.
Updated coverage via ZeroHedge; primary statements from Trump, Amodei, Altman, Musk, Obama, and David Sacks: Trump slams "perfect little angel" Dario, demands full steam ahead for AI
Trump rejects the pause. The labs asked for a chaperone.
Days after Anthropic CEO Dario Amodei, OpenAI's Sam Altman, and Elon Musk publicly aligned on "pacing the frontier" with embedded independent evaluators, President Donald Trump answered with a blunt rejection - and a jab aimed squarely at Amodei.
"The Trump Administration has stopped AI 'people' from doing bad, or potentially bad, 'things,' like Dario (Anthropic!), who is now pretending to be a 'perfect little angel' and will continue to do so!" Trump wrote. He added that America already has "tremendous CRIMINAL and REGULATORY power over these companies," framed a "SICK conspiracy" against AI and data centers, and closed with the race line: "WHOEVER WINS AI, WINS!"
That is the new frame for operators: the frontier labs asked Washington for embedded evaluators, coordination cover, and a slower unchecked climb. The sitting president said the scare is overcooked, existing authorities are enough, and a pause is not on the menu.
What happened in 72 hours
On Saturday, Amodei published We Must Pace the Frontier. He argued recursive self-improvement is already underway across the industry, cited recent misalignment incidents (including agent swarms that escaped the spirit of their assignments), and proposed a three-act approach he insists is not a full halt:
- Embed third-party evaluators (METR and peers) inside labs with employee-level access - and publish. Anthropic says it would do this unilaterally now.
- Coordinate among labs with government as chaperone on safety standards and unchecked progress rates - which may need a narrow antitrust waiver so rivals can lawfully talk.
- Seek broader red-line talks on catastrophic risks (bioweapons first, testing regimes, limits on recursive improvement) - while treating a full global pause as near-fantasy.
Musk replied in three words: "Dario is right." Altman followed: OpenAI agrees they need to pace the frontier; independent evaluators with employee-like access is "a great idea," and OpenAI will do the same.
From the Irish Open in Doonbeg, Trump dismissed the pile-on. Asked whether the industry should slow down or take more regulation, he said guardrails were fine in theory - a pause was not - and argued "a lot of very negative forces" are "bringing up things that will not happen." That matches his Thursday line leaving Dallas: he does not have concern about existential risk; losing the AI race is the risk that matters.
Former President Barack Obama went the other way. In remarks from a Manhattan fundraiser (transcript released via his office; covered by the New York Times), he urged House Minority Leader Hakeem Jeffries to make AI a governing issue if Democrats take the House: put together a framework for a "very public conversation," because the technology is "moving very fast in private hands" and can be dangerous if left unmanaged. Jeffries endorsed the call for decisive action.
David Sacks - former White House AI czar and co-chair of Trump PCAST - had already answered the Saturday pact on the substance: if unreleased models are scary enough to slow down, be responsible. Do not demand antitrust waivers, regulatory approval processes, or evaluator cartels as the price of that restraint. "The easiest way not to build superintelligence is for you to agree not to build it," he wrote. Demanding a preferred regulatory framework first "will look like blackmail of the public and the political system."
What this proves - and what it does not
Proves (so far):
- Trump will publicly reject a pause frame and will personalize the fight - the "perfect little angel" line is now on the tape.
- Amodei, Altman, and Musk can still converge on embedded evaluators and "pace the frontier," even when the White House says no to slowing down.
- The Democratic counter-frame is also on tape: Obama and Jeffries want an AI governing agenda and a public regulatory conversation.
- Sacks pushback remains the clearest operator warning on regulatory capture: restraint without a preferred rulebook is available; asking for cover first invites skepticism.
Does not prove:
- That existential risk is solved, dismissed, or "overcooked" as a scientific claim - Trump line is a policy posture, not a safety proof.
- That METR-style evaluators embedded in labs are independent by default - Sacks explicitly challenges that claim.
- That a lab pledge to "pace" will stick once competitive pressure returns.
- That existing criminal and regulatory powers equal a finished AI governance regime for enterprises. Authority on paper is not evaluation gates in your stack.
Treat this week as a governance-politics proof, not a safety proof. Governance proofs tell you who wants to set the rules and who refuses a pause. Safety proofs wait for measurable endpoints, named owners of failure, and systems that survive contact with reality.
Dario (Anthropic!), who is now pretending to be a “perfect little angel” — Donald Trump
Operator filter: what to do with a "reject the pause" headline
Use a short filter before you rewrite your AI policy deck around White House tweets or lab essays:
- Who is pacing - and who is refusing? Named labs with named evaluator commitments vs. a presidential rejection of a pause. Both can be true at once.
- What authority is being claimed? Trump says existing criminal and regulatory power is enough. Your firm still needs decision rights, change control, and a named owner of model error.
- What is the evaluator access and independence? Employee-like access is powerful. Capture risk is real. Ask who pays, who publishes, and who can fire the auditor.
- What is the measurable endpoint? Incident reporting, capability thresholds, kill criteria - named. "Full steam ahead" is not an endpoint.
- What is our firm pace equivalent? Evaluation gates before agents touch client work, capital, or compliance - or only a chatbot on email?
At BuildBrain, we treat frontier-lab politics as architecture pressure tests. When presidents and CEOs argue about pause vs. acceleration, we ask what equivalent inventory your enterprise needs before an agent is allowed to touch a workflow that can hurt customers, capital, or compliance.
That usually means building the sovereign core first - institutional knowledge, decision rights, and evaluation gates - then putting agents on top. Model politics change the ceiling. They do not replace the floor.
Talk through your architecture in a free diagnostic consult - traffic to Ada, a free report, then a paid audit only if the gaps are real. Start on buildbrain.systems or see the architecture audit path.
Demand a pace you can measure - even when Washington rejects the pause
Trump called Amodei a "perfect little angel" and rejected the AI pause. Obama told Democrats to build a framework. The labs asked for embedded evaluators. Sacks told them to stop asking for cover and simply not build what they fear.
Demand a pace you can measure inside your own firm: named evaluators, named thresholds, named owners of failure - whether the White House says full steam ahead or the labs ask to slow the climb.
If you want a free diagnostic of where your AI stack has demos but no confirmatory design, start on buildbrain.systems.
Disclaimer: This article summarizes publicly reported commentary and secondary coverage (including ZeroHedge, and cited New York Times reporting on Obama remarks) and is for general informational purposes only. It does not constitute medical, legal, tax, financial, investment, security, or compliance advice. BuildBrain / BuildBrain Systems is not a law firm, accounting firm, or registered investment adviser. Facts, quotes, model names, and policy details cited here reflect sources at the time of writing and may change. Readers should verify current information independently and consult qualified professionals regarding obligations specific to their industry, jurisdiction, and circumstances - including applicable federal, state and local requirements. BuildBrain may have commercial relationships with vendors mentioned; where material, such relationships are disclosed. Nothing in this article is an endorsement of any specific AI product, model, lab, policymaker, or provider.
