Strip out the high-flown philosophical sentiment in Anthropic’s 300 page pre-float prospectus and do what any good detective does. Follow the money.

And maybe get a bit angry when you see how its leaders think about their company and the rest of us. That’s what American voters have just done in the mid-term elections there, as the backlash against AI and data centres grows.

Reading Anthropic’s prospectus is disturbing. What might be about to become a $US1.4 trillion company desires a future that is distinctly non-human, except to the extent that we humans consume its product and produce the data and content its AI models need to eat.

Anthropic has even coined a new term for us: prosumers — creatures that produce content the company needs and consume its products while doing so. A bit like self-milking cows.

The company is selling shares to ordinary punters for billions of dollars. Most of that money will go, as fast as possible, into more and bigger data centres. Its whole business philosophy rests on models getting smarter simply because they get bigger — meaning ever more computing power, both to “train” them and to do work for paying customers (sorry: prosumers).

Chunks of the windfall go to a tiny group of very special humans: the Seven Founders, who will see their shares grow in value after the float and have given themselves a defining 50.1 per cent vote on any decisions put to shareholders. Another chunk goes to Anthropic’s actual human workforce — around 5300 people as of September 2026.

No one who buys the stock gets any of the company’s future revenue. Anthropic won’t be paying dividends like a normal public company: the cash stays in the business, spent as the Founders direct. Buyers only make money if the price rises and they sell: unlike BHP, Rio Tinto or the banks, there’ll be no dividend payments to cushion holders during bad times in a market already showing many signs of an AI bubble.

Right now, the company is spending billions on supply contracts with the likes of Amazon and Google for data-centre computing power, so those companies and their shareholders will keep getting cash out of Anthropic — for a while. If its plans come to fruition, that cashflow shrinks: Anthropic wants to own and operate the full technology stack it needs, which means eating into its suppliers’ businesses. As the prospectus says, those suppliers are also competitors and the AI price-and-value war is just getting started.

For now, though, these other hyperscale tech cos are some of Anthropic’s biggest boosters with its float. They are buying shares to ensure the $trillion plus valuation holds up, because if it didn’t, it’d send an earthquake into their own valuations and borrowing costs.

The Founders may well be important to the company’s future, so it’s great they’re being looked after. But the one group Anthropic actually depends on — we humans who create new data and content — gets none of the rivers of gold.

That content is the fuel its models run on — the food they eat to grow — and “scaling” means those needs grow fast and never stop. Yet, in a wholly counterintuitive twist that contradicts the altruistic sentiment laced through the float documentation, the company has no plans to pay any humans for what we create.

You could say “What’s new? Google and Meta have been organ harvesting our data for years” — but Anthropic’s sights are set even on data protected from that kind of harvesting by law. It is actively resisting the copyright and intellectual property laws that protect music, books, visual art and other human works.

The company wants us to admire its philanthropic gestures — those marginal distractions in the scheme of its finances, but invaluable marketing moves. The Seven Founders have magnanimously signed a voluntary but non-binding pledge to spend 80 per cent of their equity on good causes.

So money that should be paid to the humans creating the fuel its models need will be retained by the company and pump up the Founders’ bank accounts. They decide who gets a gift from their benevolence. It’s a bit like wealthy pirates using their loot to fund a few local schools, with grateful recipients meant to congratulate them on their higher values.

That gives seven people enormous power, bound only by — as Donald Trump has memorably said — their own morality. Elon Musk shows what a capricious trillionaire does with that: slowing down Starlink for Ukraine mid-war and spending US$445 million on Trump Republicans in the US mid-term elections. Nvidia’s Jensen Huang is giving us another example of the risks flowing from relying on tech bros’ judgement.

At its core, the Anthropic prospectus is a pledge to fight copyright and IP claims brought by people whose content it has harvested. The battle will be fought in the courts, and Anthropic will only pay up if it loses. More ambitiously, the company is seeking to change these pesky laws altogether so it gets its fuel for free.

Right now it has teamed up with arch competitor OpenAI to pressure the Albanese government to loosen Australia’s data protections to allow large-scale harvesting for free. The threat is that they won’t invest in data centres here unless their demands are met — cheerfully playing on national security anxieties to get a massive commercial benefit.

You have to wonder about a company that says its core mission is to create abundance, prosperity and a better future for all humanity that then plans not to share the wealth it creates with the humans whose content it needs as fuel. Relying on the limited and highly personal indulgence of its leaders to share some of the cash around on things they deem worthy is no substitute for a business model that pays humans for work done.

Human creators are essential to Anthropic’s future. So, it’s only right to expect that the company shows how it’s going to pay us for the fuel we make that it needs to survive.

That would be a human-centred AI future we humans might buy into. •

A version of this article was first published by Inside Story.