Dario Amodei on the AI Backlash: 'Fundamentally a Crisis of Trust'
Dario Amodei on the AI Backlash: 'Fundamentally a Crisis of Trust'
"I think it is fundamentally a crisis of trust. I think that ordinary people don't trust companies, governments, or the tech industry and always suspect that we are cooking up some new way to screw them over."
On Saturday, August 15, 2026, a man who mostly stays off social media wrote a long thread.
Anthropic cofounder and CEO Dario Amodei had been called out by name. Investor Gavin Baker argued — on the All-In podcast and on X — that Amodei's warnings about AI's dangers had helped fuel the American backlash against AI, especially the opposition to data centers. That he had "lost the argument" on regulation. And that as the incoming CEO of one of the world's most important companies, he owed his industry better PR.
Translation: stop scaring people, say something nice.
Amodei refused the frame. He didn't argue about whether AI is risky. He attacked the causal chain itself — the assumption that public hostility toward AI is something AI executives talked into existence. Here are the ten things worth pulling apart.
1. The Ask: Be Your Industry's Spokesperson
Baker's request was unusually explicit:
"he is about to be the CEO of one of the most important companies in the world," Baker wrote, "I respectfully think he should make an effort to be a more positive advocate for his own industry."
Look at the premise smuggled into that sentence: brand management is part of the job description. And Baker isn't alone. Plenty of people argue that AI skepticism and government crackdowns are the natural consequence of executives issuing dire warnings — there's even a case that Anthropic's own safety warnings backfired into government action against its most powerful model.
This isn't a technical disagreement. It's a disagreement about what a frontier lab CEO is for: industry advocate, or risk discloser.
My take: The trouble with "say something nice" is that it assumes public sentiment is a messaging problem. If the public were that easy to manage, advertising agencies would run the world.
2. "My Messaging Isn't Actually Lopsided"
Amodei's first move was to reject the characterization:
Amodei disagreed with the idea that his "messaging has been disproportionately negative." Instead, he said that his writing has been "about equally balanced between risks and benefits," and that he wrote his essay "Machines of Loving Grace" because he "didn't feel the AI industry was painting an inspiring enough picture of how the technology could radically transform the world for the better."
He added a pointed complement to that: the clips that make him look gloomy are cut for clicks on social media.
The defense holds up. Machines of Loving Grace imagines compressing a century of biomedical progress, ending mental illness, and eliminating extreme poverty — optimism bordering on science fiction. Nobody shared that essay. What they shared was a fifteen-second clip about AI wiping out half of white-collar jobs.
My take: A person's full position and their algorithmic silhouette are two different objects. He controls the first. The recommendation engine controls the second. Complaining about it changes nothing — but criticizing the silhouette as if it were the position isn't fair either.
3. The Real Diagnosis: This Was Never About AI
Here's the actual fault line. Amodei concedes the conclusion and rejects the attribution.
He agrees the public has a negative view of AI. He agrees it's a big problem. What he rejects is the claim that this negativity was "primarily caused" by him, or by any AI leader warning about risk.
"I think it is fundamentally a crisis of trust," Amodei said. "I think that ordinary people don't trust companies, governments, or the tech industry and always suspect that we are cooking up some new way to screw them over."
In his telling, that crisis has been decades in the making, and the AI backlash is "just the latest iteration of it."
My take: This is the most load-bearing sentence in the whole exchange. Draw a line through the last twenty years of tech narratives: social media promised connection and delivered filter bubbles; the gig economy promised freedom and delivered precarity; crypto promised decentralized finance and delivered a crater. By the time AI walked onstage, the audience wasn't a blank slate. It was a room full of people who'd been sold the same pitch three times already.
4. The Criticism He Thinks He Actually Deserves
Amodei volunteered a sharper attack on himself — naming Anthropic explicitly:
"I think by far the most accurate criticism of AI companies including Anthropic is that we haven't yet delivered on our big promises to benefit the world," Amodei said. "That is totally on us, and I think it's the criticism you should be making, instead of all this stuff about messaging and marketing."
It's a rare move: the person under attack handing over a more powerful weapon than the one being used on him.
My take: There's real politics in this. It drags the fight from "how you talk" back to "what you've shipped" — and on that ground, Anthropic is at least standing at the same starting line as its critics. Honest, yes. Also a very deft change of subject.
5. "Curing Cancer" Has Been Inflated Into Worthlessness
Following that thread, Amodei lands one of the sharpest observations in the exchange:
he said promising that AI will cure cancer is "more a cliche than it is inspiring;" what would actually change people's opinions of AI would be "actually curing cancer."
Fortune records an even harsher version of the judgment: an ad claiming AI will cure cancer would most likely be dismissed as deceptive on sight.
Anthropic is pushing harder into biology and medicine, Amodei says, and there are "early glimmers" of what could be extraordinary results. But until there's something real, he's done making promises.
My take: The marginal utility of a promise decays into negative territory. The first time you hear "AI will cure cancer," it's hope. The tenth time, it's noise. The hundredth time, it's a tell that someone's selling you something. That inflation curve was printed by the AI industry itself.
6. "Regulation = Regulatory Capture = Concentration of Power" Is Silicon Valley Shorthand
On regulation, Amodei said Baker was constructing a false choice: either distribute AI widely without regulation (including via open models), or regulate and concentrate the technology in a handful of companies.
"I know that there's a sort of Silicon Valley shorthand where regulation = regulatory capture = concentration of power, but I've always found this to be an overly simplified picture of the world," Amodei said. "Many people outside this bubble think of regulation as something that constrains corporate power and benefits ordinary people."
He points to the court system as an institution that decentralizes power. Then — interestingly — he says he doesn't necessarily agree with the outside-the-bubble view either, which is why Anthropic writes its policy proposals so carefully:
"We try very hard to make proposals that disadvantage (slow down) frontier AI companies while advantaging smaller competitors."
My take: The equation "regulation inevitably entrenches incumbents" has been treated as an axiom in Silicon Valley for so long that nobody re-derives it anymore. Sometimes it's true. Sometimes it isn't. It depends entirely on how the rules are written. Promoting an empirical observation to a law of physics is the laziest possible way to win an argument.
7. The Concentration Comes From Scaling Laws, Not From Regulators
This is the technical foundation of his whole case, and he gives his opponents half of it:
Amodei said that in his view, "AI is structurally a technology that tends to concentrate power"
But he attributes that to scaling laws — model performance improving with the resources poured into it — rather than to regulation.
The attribution does real work. If concentration comes from regulation, the fix is deregulation. If it comes from scaling laws, deregulation fixes nothing, because compute and chips were never evenly distributed to begin with.
My take: This is the pivot the whole argument turns on, and it's the easiest step to skim past. Conceding "AI naturally concentrates power" looks like handing ammunition to the other side. It's actually a change of terrain: if concentration is physical rather than regulatory, the question stops being whether to regulate and becomes what mechanism offsets a structural tilt.
8. The Line Read as a Declaration of War on Open Source
Here's the sentence that detonated:
"open-weights do help some with this but are nowhere near a sufficient solution because they simply shift the concentration somewhat to those with the most compute and chip."
Figures in the tech world, David Sacks among them, read statements like this as a declaration of war on open-source AI.
Read the actual words and there are three claims stacked in there: open weights do help some, they're nowhere near sufficient, and the reason is that concentration merely relocates. He does not say ban them — the very next passage explicitly carves out room for open-weight models.
That said, let's be clear about where everyone is sitting. Anthropic is the flagship of the closed-model business. Its verdict that open weights are "not sufficient" points in a convenient direction.
My take: The critique itself is strong — publishing weights doesn't help much if you still need tens of thousands of accelerators to serve them, so power moves from whoever owns the model to whoever owns the compute. But it loses force coming from a closed lab. The argument can be correct and the messenger can be interested. Both are true at once.
9. What He Proposes Instead: 'Rules of the Road'
Having rejected the binary, Amodei offers a three-target framework:
"By contrast I think the right 'rules of the road' can simultaneously (a) address AI's cyber/bio/alignment risks, (b) institutionally constrain the power of the frontier AI companies, and (c) leave room for open-weights models while also addressing the specific risks that they bring."
Concretely, he says he supports creating a FINRA-like entity — the self-regulatory body model from finance — and backs the Trump administration's stance on AI testing.
Note item (b): institutionally constrain the power of the frontier AI companies. He runs one.
My take: The ambition here is that it refuses every ready-made camp — not laissez-faire, not blanket restriction, and it deliberately leaves a door open for open weights. That's also the problem. A rule that hits all three targets at once is brutally hard to draft and harder to implement. Beautiful framework. The devil is in the statutory text.
10. The Closing Move: Honesty as Strategy
Amodei's final line pulls the whole thing out of PR and into method:
"Honesty is the right thing on the merits, and in terms of public credibility and trust it is no worse than, and may in fact be better than, an approach that ignores or distracts from risks which people instinctively understand are real."
This answers Baker directly. You want me to accentuate the positive — but the public already knows the risks are real. Pretending otherwise, when they already know, doesn't build trust. It accelerates the collapse of it.
My take: This is the sturdiest claim in the thread, because it doesn't rest on Amodei's character. It rests on a checkable fact: the public has gotten harder to fool. In an industry where everyone is over-promising, telling the truth may genuinely be the strongest differentiated position available. Provided, of course, that he delivers on point 4.
The Bottom Line
On the surface this is a fight about how AI executives should talk. Underneath it's a harder question: once an industry has overdrawn its credibility, what can it possibly spend to earn it back?
Baker's answer is narrative. Reframe it. Tell a better story. Amodei's answer is delivery. Stop talking. Go cure the cancer.
The striking thing is how much they agree on. Both accept that the public distrusts the AI industry, and both call it a serious problem. The entire dispute is over whether to book that liability under "said the wrong thing" or "didn't do the thing."
And the standard Amodei sets for himself is the most dangerous of any response available: he's staked Anthropic's credibility on producing real results. There's no PR shortcut on that road. Just a binary outcome — deliver, or get pinned to your own words a few years from now.
Sources: TechCrunch: Anthropic CEO says AI backlash is 'fundamentally a crisis of trust' Fortune: Dario Amodei admits AI suffers from a crisis of trust