Trump criticises AI boss who called for slowdown and says only guardrail needed is a 'smart president'
Author: Gammatek ISPL, Published: Sep 2026
In this article
What Trump actually said, and what triggered it
The Anthropic dispute behind the personal dig
Why Congress isn't going to fill the gap
The part that matters more than the tweet: state law didn't pause
Federal rhetoric vs. enforceable reality, side by side
Implementation considerations: building guardrails that don't depend on who's president
FAQ
What Trump actually said, and what triggered it
On Monday, September 14, following a weekend in which a group of AI industry leaders publicly called for a coordinated slowdown in development, Trump posted on Truth Social that AI needs no formal oversight beyond executive judgment. His exact words: "The only control or 'guardrails' that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!" He added that his administration already holds "tremendous CRIMINAL and REGULATORY power" over AI companies and accused unspecified critics of running a "SICK conspiracy" against AI and data centers, framing the entire slowdown movement as something that only benefits China.
The post was a direct response to a weekend of what one outlet called "shocking unity" among AI leaders, several prominent AI CEOs, including Anthropic's Dario Amodei, publicly backed calls to pace development more cautiously, warning about risks ranging from loss of control over increasingly capable systems to broader societal disruption. The reaction was immediate and measurable: AI-linked stocks fell in early trading Monday as investors weighed the possibility that a genuine industry slowdown could materialize, even though, at time of writing, nothing about the actual regulatory environment had changed. The Nasdaq remains up roughly 12% year to date and the S&P 500 up more than 10% over the same period, so the stock dip reads as a sentiment reaction to slowdown talk rather than a response to any policy that's actually been enacted.
The Anthropic dispute behind the personal dig
Trump's post singled out Amodei by name, calling him "Dario (Anthropic!), who is now pretending to be a 'perfect little angel.'" That's not a generic jab, it follows a specific prior conflict. Anthropic reportedly declined to remove certain safety restrictions from its models for Department of Defense use, after which Trump ordered the federal government to phase out its use of Anthropic's products. That history matters for reading Monday's post accurately: this isn't a president discovering AI safety concerns for the first time and rejecting them in the abstract. It's a continuation of a specific, ongoing dispute between the administration and one AI lab over how much safety restriction is appropriate on government contracts, now generalized into a broader public statement about regulation overall.
For a business audience, the useful takeaway isn't the personal dynamics, it's what the underlying dispute reveals: this administration's posture on AI safety guardrails has already resulted in one concrete government action (the Anthropic phase-out), which is proof that policy actually moves in specific, uneven ways based on individual disputes, not through a stable public framework a business could plan compliance around in the first place.
Why Congress isn't going to fill the gap
If federal executive action is unpredictable, federal legislative action looks unlikely at all. Speaker of the House Mike Johnson pushed back directly on the idea that Congress should move quickly, telling CNN that "if Congress just races in and does some sort of emergency session to try to regulate AI, we will lose the race to China, and that is a threat to every single American." Vice President JD Vance struck a more measured note the same morning, saying the administration remains "concerned" about AI and "thoughtful about the risk," even while maintaining the priority of staying ahead of China, a small but real daylight between the vice president's framing and the president's, though not enough to suggest legislative movement is imminent.
This isn't a new dynamic; it's a continuation of a deregulatory push the administration has been pursuing for months. In December 2025, Trump signed an executive order titled "Ensuring a National Policy Framework for Artificial Intelligence," explicitly designed to preempt state AI laws the administration considers inconsistent with federal policy, naming Colorado's AI Act directly as a target. The stated rationale was that a "patchwork" of state-by-state rules burdens innovation more than a single national standard would. Congress, for its part, already weighed in on this fight once: a proposed federal moratorium that would have frozen state AI legislation was stripped from the One Big Beautiful Bill Act by a 99-1 Senate vote in July 2025, an almost unanimous rejection of federal preemption that undercuts the idea that Washington is on the verge of stepping in with a unified national framework any time soon.
Put plainly: the executive branch wants fewer rules and to override the states that have written their own. The Senate, in the one moment it had a clean vote on the question, overwhelmingly said no to freezing state action. That's not gridlock resolving in either direction, it's gridlock, and gridlock doesn't produce a compliance framework a business can rely on.
The part that matters more than the tweet: state law didn't pause
Here's the detail that gets lost in coverage of the Trump-Amodei feud: while the federal fight over whether to regulate AI has dragged on, states kept legislating anyway, and several of those laws are already binding. As of mid-2026, at least five states, California, Colorado, Texas, New York, and Illinois, have comprehensive AI laws either already in force or set to take effect by January 1, 2027, layered on top of more than 40 narrower state laws covering deepfakes, hiring tools, and AI chatbot disclosure specifically. State legislatures introduced 1,208 AI-related bills in 2025 alone, up from 635 the year before and fewer than 200 in 2023, a more than fivefold increase in two legislative cycles.
Some of the specifics that apply right now or within the next several months:
Colorado replaced its original comprehensive AI Act with SB 26-189 in May 2026, a narrower law regulating "automated decision-making technology" used in consequential decisions about employment, housing, healthcare, and similar categories, effective January 1, 2027. A companion law, the Chatbot Safety Act, requires operators to disclose AI interaction to users and build in safeguards for minors, and formal rulemaking on both was underway with public comment open through late October 2026.
California has gone further than any other state in sheer legislative volume, including its AI Transparency Act, which requires disclosure and detection tooling for AI-generated content, and new Civil Rights Council regulations governing AI use in employment decisions, which took effect October 1, 2025, and explicitly make the presence or absence of bias testing relevant evidence in discrimination claims.
Illinois requires employers to disclose AI use in employment decisions under a law effective January 1, 2026.
New York City's Local Law 144, requiring bias audits for automated employment decision tools, has been in force since 2023 and remains active regardless of anything happening at the federal level.
There's also an active legal wrinkle worth knowing about if you operate in Colorado specifically: a federal court enjoined enforcement of the original Colorado AI Act after xAI and the Department of Justice jointly sought a preliminary injunction, which is a live illustration of exactly how unsettled this area is even within a single state. The practical lesson isn't "ignore state law because it's contested," it's "assume state law applies until a court says otherwise, because betting on litigation outcomes is not a compliance strategy any auditor will accept after the fact."
Federal rhetoric vs. enforceable reality, side by side
Stripping out the political noise, the actual compliance landscape splits cleanly into what's rhetoric and what's enforceable today. This is the distinction worth internalizing, because conflating the two is exactly how a company ends up unprepared.
Signal | What was said or proposed | Is it enforceable against your business right now? |
Trump's "smart president" Truth Social post | AI needs no guardrails beyond executive judgment | No. It's a rhetorical statement, not a rule, regulation, or law |
December 2025 executive order on state preemption | Federal policy should override state AI laws deemed inconsistent with it | Partially, contested. Preemption is being litigated state by state, not settled uniformly |
Senate's 99-1 rejection of the federal AI moratorium | States should retain authority to regulate AI independently | Yes, in effect. It confirms state laws remain the operative rules unless a specific court overturns them |
Colorado SB 26-189, Illinois HB 3773, NYC Local Law 144, California employment AI rules | Specific disclosure, bias-testing, and human-review obligations for employers and deployers | Yes, currently in force or scheduled with firm effective dates |
Speaker Johnson's comments on Congressional inaction | Congress shouldn't rush emergency AI regulation | Not a rule, but a strong signal that no federal statute is coming soon to replace or clarify the state patchwork |
Read across that table, and the pattern is unambiguous: every genuinely enforceable obligation in this landscape right now is coming from the states, not from anything the president has said or proposed. Building a compliance plan around federal rhetoric, in either direction, treating a dismissive tweet as license to do nothing, or treating a preemption executive order as a green light to ignore state law, is a bet on political outcomes that a compliance program shouldn't be making.
Implementation considerations: building guardrails that don't depend on who's president
If Monday's news changes anything about your actual risk exposure, it should be a prompt to check whether your AI governance work has been implicitly waiting on federal clarity that isn't coming. A few concrete steps worth taking regardless of how the Trump-Amodei dispute or the broader slowdown debate plays out:
Map where you operate against the current state AI law list, not the federal one. If you have employees, customers, or applicants in California, Colorado, Illinois, New York, or Texas, you likely already have binding obligations, independent of anything happening in Washington this week.
Treat "consequential decisions" as the trigger, not "we use AI." Every one of the major state laws targets AI used in hiring, lending, insurance, housing, healthcare, or similarly high-stakes categories specifically. A low-stakes internal drafting tool and a hiring-screening algorithm are not the same compliance question, and lumping them together either overspends on the low-risk tool or underprepares for the high-risk one.
Build the traceability and human-oversight practices once, and let them satisfy multiple regimes at once. The accountable-owner and audit-trail principles we've covered in our piece on OpenAI's own concentration-of-power warning map directly onto what Colorado's ADMT rules and NYC's Local Law 144 bias-audit requirement both expect: a named, responsible party and a documented review process. You don't need a separate framework per state if the underlying practice is solid.
Don't treat federal deregulatory signals as a reason to relax employment-related AI compliance specifically.Hiring and employment AI is the fastest-growing enterprise compliance category at the state level right now, and it's the category with the most direct litigation exposure if it goes wrong, regardless of the federal mood.
Reassess quarterly, not annually. Colorado alone rewrote its comprehensive AI law once and delayed it twice within about eighteen months. A compliance posture built to be revisited once a year will be stale before the year is over in this environment.
That instability is exactly why we've been building this cluster around the assumption that political rhetoric and regulatory risk move on different timelines: the trust and accountability practices we covered in why the AI industry's trust problem is now every vendor's problem and the traceability principle from OpenAI's own policy team both hold up regardless of which way the federal political winds blow this month, because they're built around what auditors and state regulators actually ask for, not around who currently occupies the White House.
Frequently asked questions
Does Trump's statement about AI guardrails have any legal effect on businesses? No. It's a social media post expressing a policy preference, not an executive order, regulation, or statute. It carries political signal but no enforceable weight on its own. The December 2025 executive order aimed at preempting state AI law is the closer thing to an actual policy action, and even that is being contested in court rather than settled.
If the federal government wants to preempt state AI laws, should businesses stop complying with them? No. Preemption is being litigated on a state-by-state, law-by-law basis, not resolved uniformly. Treating a proposed federal preemption as an excuse to disregard an existing, unrepealed state law is a bet against pending litigation, which isn't a defensible compliance position if a state regulator or plaintiff's attorney disagrees before a court rules.
Which states have the most urgent AI compliance requirements for businesses right now? California, Colorado, Illinois, New York, and Texas currently have the most developed frameworks, with employment-related AI use, hiring tools, automated decision-making, and bias-audit obligations, representing the fastest-growing and most litigation-exposed category across all five.
How does the Anthropic-Trump dispute affect companies that use Anthropic's or other AI vendors' products?Directly, if you sell into or rely on federal government contracts. The administration's decision to phase out Anthropic products from federal use is a live example of how quickly vendor risk can shift based on a specific policy dispute, which is itself an argument for building vendor-agnostic AI governance practices rather than assuming any single vendor relationship is stable long-term.
Is a coordinated AI industry slowdown actually going to happen? It's genuinely uncertain. Multiple prominent AI CEOs backed the idea over the weekend, which moved markets, but the administration has shown no interest in mandating one, and the industry's own history of voluntary coordination on safety measures is mixed. Businesses shouldn't plan around either outcome and should instead focus on the compliance obligations that are already enforceable regardless of whether a slowdown materializes.
What's the single biggest mistake a business could make in response to this news? Reading it as reassurance that AI regulation isn't coming and can be deprioritized. The state-level data says the opposite: AI compliance obligations have grown more than fivefold in two years and show no sign of slowing, irrespective of the federal rhetoric.
Not sure whether your AI-assisted hiring, customer service, or decision-making tools would satisfy Colorado's SB 26-189, Illinois's disclosure law, or California's employment AI regulations if a regulator asked tomorrow?
Our AI governance and compliance review maps your current AI use against the specific state laws that apply to where you operate, identifies which tools count as "consequential decision" systems under each framework, and builds the documentation you'd need to demonstrate compliance regardless of what happens at the federal level.
Comments