AI Billionaires Want to Control EVERY Aspect of Your Life

ELI5/TLDR

Karen Hao, who wrote a book on OpenAI called Empire of AI, returns a year later to argue that the AI boom is less a technology story than a power story: a handful of companies trying to make themselves the people who decide everyone’s future. She walks through the physical underside of it all — the mines, the water, the power plants, the cheap overseas labour — and the political machinery on top: super PACs, planted lobbyists, friendly ex-prime-ministers. Her core claim is that the industry’s favourite line, “this is inevitable, nothing you can do,” is itself a tactic to make people give up. And her hopeful counterpoint is that, over the last year, people stopped buying it.

The Full Story

From “nothing we can do” to people pushing back

A year ago, when Hao first toured with the book, the mood she met everywhere was helplessness. People felt the AI rollout was a weather system — something that happened to them. Her argument now is that this feeling isn’t an accident. It’s the product.

“The most important way to understand the AI story today as it’s conceived of by Silicon Valley is that it is a political project. And the central feature of this political project is taking agency away from people.”

The good news, in her telling, is that the helplessness is cracking. She has launched a project called the AI Resistance List documenting around 30 pushback efforts worldwide. One favourite: a community in Quilicura, Chile, sitting in one of Latin America’s data-center hotspots, built a website that looks like a chatbot — but when you type a question, a real local person answers. Ask for a picture of a dog and a local artist named Benji draws one on paper, photographs it, sends it back. The point is to make a buried fact visible: the “cloud” is not ethereal. It runs on someone’s land, water, and electricity.

The fashion-industry analogy

Bastani presses the obvious hypocrisy — people criticise AI and then use it for, say, planning a no-fly beach holiday in Normandy. Hao refuses to play the gotcha. Her frame is the garment trade.

“The answer was not no one wear clothes anymore because it’s unethical to wear clothes. The answer was these clothes are being developed in a really deeply unethical way.”

Sweatshops weren’t fixed by everyone going naked. They were dented by labour organising, consumer backlash, regulation, and the eventual creation of an “ethical” market that brands could charge a premium for. She wants the same arc for AI: keep the genuinely useful tools, but force the supply chain — the data, the labour, the energy — onto less corrosive foundations.

That supply chain is real and ugly. The Democratic Republic of Congo, source of cobalt and other minerals, sits directly in the AI hardware chain, with child labour and environmental wreckage attached. The point she keeps returning to: a glass screen hides a very physical, very extractive reality.

”End times fascism” and the bunkers

The interview’s most striking idea, borrowed from Naomi Klein, is “end times fascism.” Hao says she has heard senior Silicon Valley figures agree with the first half of her own analysis — yes, AI as currently built will worsen inequality. The split is what comes next. She concludes: so let’s stop. They conclude: it’s inevitable, so make sure we’re among the winners.

“There’s no point in trying to say… the end is nigh… and they don’t have any desire to save what we have currently. They’re just worried about getting to the next place and making sure that they are okay.”

Hence the bunkers. She’d been in New Zealand and learned why the billionaire set picked it: model the blast radius of a nuclear exchange and New Zealand mostly sits outside it. Believing in the end of the world is convenient if you’ve already bought a seat outside it — it removes both moral responsibility and personal consequence in one move.

Bastani brings in Neil Postman’s Technopoly (societies that end up serving their tools rather than the reverse) but Hao’s emphasis is more material than philosophical: the fatalism is convenient because it’s profitable. It’s the same machinery now minting the first trillionaire — Elon Musk, via the coming SpaceX IPO that bundles in xAI.

The “three wise men”: Altman, Musk, Amodei

Hao is sharpest on the personalities.

On Sam Altman: his superpower was persuasion — of investors, of employees, of the public. As an investor he never had to make hard trade-offs because he could always raise more money, so one theory is he simply never built the muscle of choosing. Inside OpenAI, sources described the culture in one word: chaos. Going public will force some governance, but she’s clear it won’t fix the underlying problems — Google is already public and arguably has more leverage over daily life. The bigger shift, she says, is that Altman has lost the narrative. Reality caught up — layoffs, data-center neighbours — and a man whose power was persuasion now looks defensive.

On Elon Musk and data centers in space: she’s deflating. Could you put servers in orbit? Technically there already are some. But the pitch (free cooling, no land) doesn’t survive contact with physics or economics. Maintenance would be brutal, and AI training hardware is already so unstable that researchers can’t always tell whether a weird model result is a breakthrough or a failed chip. Put that flaky kit somewhere you can’t reach it and you’ve made debugging impossible. Her verdict: it’ll go the way of the Boring Company — one underwhelming demo dressed up as the future.

On Anthropic, the supposed “good guys”:

“To me OpenAI is coal and Anthropic is clean coal.”

Same labour exploitation, same extraction, same scale-at-all-costs ideology — which makes sense, she notes, because Dario Amodei carried that playbook over from OpenAI. The damning detail: Anthropic recently partnered to use Musk’s Colossus supercomputer in Memphis, the facility that ran 35 unlicensed methane gas turbines in a community already carrying a cancer rate four times the national average.

The politics: super PACs, planted questions, friendly ex-PMs

The back half turns to the lobbying machine. Having learned from crypto, the AI industry is now playing electoral hardball because it’s losing public opinion. A super PAC called Leading the Future has amassed over $100 million (Andreessen Horowitz and OpenAI’s Greg Brockman among the big cheques) and openly states it will boost allies and destroy opponents. Alongside the visible money runs dark money — a Wired report by Taylor Lorenz described detailed scripts handed to social-media influencers, down to “film yourself in the kitchen getting the kids ready for school while talking about how much you love using AI,” plus a steady drumbeat about scary Chinese AI to justify more American AI.

The mindset underneath is quasi-religious. Brockman called donating to Trump “the most moral thing I can do.” If you believe advancing AI is the supreme good, then anyone slowing it is committing a sin.

“That’s religion, isn’t it?… It’s I’m good, you’re evil, you’re a sinner, we’re the anointed.”

It’s global, too. In the UK, Tony Blair keeps amplifying the “AI is inevitable” line while his institute takes large sums from Oracle’s Larry Ellison. In Brussels, Hao recounts lobbyists not just visiting offices but planting people in the audience at public events to ask pre-scripted questions and normalise a frame — a tactic Bastani drily compares to Bolshevik agitators seeding Soviet meetings before 1917. The EU, in a fit of “why don’t we have our own tech giants” self-doubt, has reportedly discussed loosening GDPR to allow personal data for AI training, and a tech lobby group apparently slipped language into a draft bill granting corporate secrecy over data centers’ environmental impact.

Hope, deliberately chosen

Hao admits she finished the book depressed and had to actively climb out. The book that helped was Rebecca Solnit’s Hope in the Dark, whose argument she paraphrases: real moral progress comes from ordinary people and grassroots movements, never from elites (who like the world that made them elite), and you do the work without expecting to see the harvest yourself.

She closes with a concrete payoff. OpenAI recently shuttered its video tool Sora, and every reason traces back to grassroots pressure: over $100 billion in data-center projects stalled by protests in 2025; a shaky balance sheet ahead of IPO as Wall Street starts pricing in the public backlash; and flatlining usage from consumer boycotts. A small linguistic jab — “microslop” for Microsoft’s AI output — sits in the same toolkit as the artist tool Nightshade, which subtly poisons images so that any model scraping them without consent degrades. Small acts, she insists, compound.

Key Takeaways

Hao’s central thesis: AI is best understood as a political project whose goal is concentrating decision-making power, and “it’s inevitable” is a deliberate tactic to induce surrender.
The AI supply chain is physically extractive — Congolese mining (with child labour), water, land, and power — hidden behind a clean glass interface.
Her preferred model for change is the fashion industry: not abstinence, but organising, regulation, and building an ethical alternative market.
“End times fascism” (Naomi Klein): elites who concede AI worsens inequality but treat collapse as inevitable and simply plan to be among the survivors — hence the New Zealand bunkers, sited outside likely nuclear blast radii.
Elon Musk is on track to become the first trillionaire via the SpaceX IPO, which bundles xAI.
Sam Altman’s power rested on persuasion; he has now “lost the narrative” as real-world AI impacts contradict his promises; OpenAI’s internal culture was repeatedly described as “chaos.”
Data centers in space (Musk’s pitch): physically and economically dubious — cooling claims don’t hold, maintenance is impractical, and current AI hardware is too unstable to debug remotely. Likened to the Boring Company.
Anthropic = “clean coal”: same extractive playbook as OpenAI (Amodei carried it over), recently partnered to use Musk’s Memphis Colossus, which ran 35 unlicensed methane turbines near a community with 4x the national cancer rate.
Going public won’t fix OpenAI’s core problems — Google is already public and arguably more entrenched, using search dominance to funnel users into its AI.
The super PAC Leading the Future has raised $100M+ and openly threatens opponents; dark-money campaigns script influencers and stoke “scary Chinese AI” fears.
Greg Brockman framed donating to Trump as “the most moral thing I can do” — a quasi-religious view of AI progress as supreme good.
Lobbying is global: Blair amplifying “inevitability” while the Tony Blair Institute takes Oracle money; lobbyists planting scripted questioners in Brussels audiences; reported EU moves to weaken GDPR and grant data-center secrecy.
Evidence the backlash works: OpenAI shuttered Sora amid $100B+ of protest-stalled data-center projects, a souring IPO narrative, and flatlining usage.
Tools of resistance range from the linguistic (“microslop”) to the technical (Nightshade, which poisons scraped training images).

Claude’s Take

Two things are true at once here, and it’s worth holding them apart.

The reported substance is solid, and Hao is a real journalist, not a pundit. The factual spine — the Memphis turbines, the Stargate podium lineup, the Leading the Future PAC, the Taylor Lorenz influencer-script story, the Sora shutdown — is checkable and largely checks out. Her best contribution is dragging the conversation from the abstract (“will AI take my job”) to the concrete and physical: who lives next to the data center, where the cobalt comes from, what the grid actually burns. That materialist lens is genuinely clarifying, and the “you don’t have to choose between using the tool and criticising the supply chain” framing is more honest than most takes from either camp.

Now the framing filter. This is Novara — left-wing media — and the conversation is a friendly one between people who already agree, so the contrary case never gets a real seat. “Empire,” “the generals,” “imperial project” is vivid but does a lot of persuasive work; calling something an empire is not the same as demonstrating coordinated intent, and a lot of what she describes is better explained by ordinary competitive capitalism and herd behaviour than by a unified plan. The “clean coal” line about Anthropic is a great soundbite that flattens real differences in safety posture into zero — convenient for the thesis, but a claim, not a finding. And the closing optimism, while pleasant, leans on motivated reasoning: Sora was almost certainly killed mostly by compute economics and weak monetisation, with protest as a contributing rather than decisive factor. The “it’s all grassroots” reading is the mirror image of the “it’s all inevitable” fatalism she’s attacking — both are stories people tell to feel a certain way.

Net: a high-signal interview if you can separate the reporting from the rhetoric. The facts are worth knowing; the cosmology around them is optional. A 7 — substantive and well-sourced, docked for being a one-sided room and for soundbites occasionally outrunning the evidence.

AI Billionaires Want to Control EVERY Aspect of Your Life | Aaron Bastani Meets Karen Hao