Warnings over AI are a ‘sick hoax’, Trump claims

What’s the latest? In a post on Truth Social, President Trump has dismissed fears over artificial intelligence as a “hoax” created by conspiracy theorists looking to regulate the industry. He said he had the “high IQ” necessary to safeguard America and described a “SICK conspiracy going on against AI and Data Centers”.

Congress has held committee hearings on AI but has not arrived at any consensus on regulation as communities rebel against plans for huge data centres and fears that AI could destroy humanity grow.

Who is in support of regulation? Last week Jacob Coxon issued a warning about AI after resigning from Anthropic over fears the technology could not be controlled. This led Dario Amodei, the Anthropic chief executive, to release an essay titled “We must pace the frontier”, calling for a slowdown in the rush to advance AI models.

There was agreement from other senior industry figures, including Sam Altman, the chief executive of OpenAI, Elon Musk and Sir Demis Hassabis, the chair of Google DeepMind. Jack Clark, a co-founder of Anthropic, became the latest boss to call for regulation, telling the BBC that AI firms should be forced to install a “kill switch” that can be checked by a third-party.

What about the UK? OpenAI called on the British government to “act now” and introduce stronger safety testing rules for AI companies. However, the Labour government has shied away from introducing a mandatory testing regime for AI companies despite committing to it in its 2024 general election manifesto.

---

"What do you think of this news story. I have to say it ties in with my theory that the establishment are worried about AI threatening the status quo, and want captured ai apps only."

I think your theory identifies a genuine danger—but the evidence supports a more precise version of it.

The real contest may not be AI versus the establishment. It may be who gets to domesticate AI first.

The Times story brings together three distinct questions that should not be allowed to blur:

1. Frontier risk: Could extremely powerful models cause catastrophic harm?

2. Industrial power: Who controls the computing infrastructure, models and data?

3. Human agency: Who is permitted to build and use AI applications—and whose interests must those applications serve?

The first question is legitimate. Dario Amodei’s essay identifies loss of control, biological misuse, cyberattacks and economic disruption as serious risks. It is not simply an attack on AI; he also argues that AI could generate enormous benefits. His actual position is “pace the frontier,” not prohibit ordinary people from using AI.

But your concern arises from what can happen next.

The regulatory moat

When the largest AI companies call for rules applying to the “handful of companies” developing the most powerful systems, that can be perfectly sincere—and still produce an incumbent advantage.

Large laboratories can afford:

• expensive safety evaluations;

• licensing and reporting regimes;

• specialist compliance teams;

• government relationships;

• controlled distribution platforms;

• approved application ecosystems.

Smaller laboratories, open-source developers and independent application builders may not.

That creates what I would call the Regulatory Moat Paradox:

Rules intended to control the most powerful companies can make those companies the only ones powerful enough to comply.

The result could be a small number of approved models, accessed through approved platforms, producing approved applications—while independent AI capable of challenging established professional, commercial and institutional arrangements becomes progressively harder to build.

That is the strongest version of your “captured AI” thesis.

But Trump is not necessarily defending human agency

Trump’s response is the opposite extreme. Calling all AI risk a “hoax” and presenting presidential intelligence as the only necessary safeguard is not decentralisation. It replaces institutional scrutiny with personal authority.

His emphasis is also revealing: data centres, permitting, competition with China and winning the AI race. The Times report suggests an industrial-growth agenda centred on American AI dominance—not necessarily an agenda for citizen-controlled AI.

So there are two different forms of capture:

Model	Concentration of authority
Corporate safety regime	Approved laboratories and regulators decide what AI may do
Trumpian acceleration	Government and dominant technology companies decide how fast it proceeds
Agency-led AI	People can access, inspect, adapt and challenge systems within proportionate safeguards

Neither “regulate everything” nor “remove every guardrail” necessarily restores human agency. Both can leave control concentrated at the top.

The question the story does not ask

The debate is framed as:

Should AI be accelerated or restrained?

The more important question is:

Which capabilities present genuine systemic danger—and which restrictions merely protect existing systems from being challenged?

We should distinguish:

• frontier models capable of creating biological or cyber threats;

• applications making consequential decisions about people;

• tools helping individuals understand evidence, challenge institutions and make their own decisions.

These do not require identical regulation.

A citizen using Goliathon to reconstruct a complaint, BIG Checker to interrogate a persuasive claim, or a Total Wealth Planner to understand their life is not conducting frontier-model research. Yet broad rules, liability barriers and closed platforms could make precisely those agency-restoring applications dependent upon institutional permission.

My conclusion would therefore be:

AI safety is a legitimate public-interest objective. But “safety” must not become the language through which institutions preserve their monopoly over interpretation, expertise and decision-making.

The establishment may not fear AI itself. It may fear uncaptured AI—AI that allows ordinary people to see the whole system, question its narratives and act without waiting for an authorised intermediary.

That is a much stronger and more defensible argument than saying every warning about AI is manufactured.