Did Amazon Quietly Sic the Government on Anthropic? Inside the AI Power Game

The scariest phrase in AI this week isn’t “government regulation.” It’s “the hand that moved the government.” Last week, US officials abruptly cut off access to Anthropic’s Fable 5 and Mythos 5. You probably saw it. But one name keeps surfacing in the conversation about what was really behind it: Amazon CEO Andy Jassy.

Let me be upfront. The “Amazon was behind it” theory is, for now, speculation and circumstantial reading from the community and the industry. It is not a confirmed fact, and I won’t pretend it is. But why the suspicion landed this fast, and what the underlying structure tells us, is worth taking seriously.

Start With the Timeline

Lay out the last several days in order and the picture sharpens.

First, Anthropic issued a public apology over allegations that it had baked invisible guardrails and distillation traces into its models. Almost immediately after, the US government ordered a suspension of access to Fable 5 and Mythos 5. Then an “open source AI must win” movement erupted across developer communities.

Read at face value, this looks like an ordinary sequence: a quality scandal leads to regulation. But one question lingers. Why now, why this fast, and why this hard?

Why the “Amazon Theory” Even Exists

The core of the suspicion is simple: Jassy reportedly spoke with US officials, and that conversation is said to have triggered the government’s hardline move against Anthropic’s models.

On the surface, that makes no sense. Amazon is one of Anthropic’s largest investors. Why would a company that poured in billions turn around and have the government beat up its own portfolio company? That contradiction is exactly what makes the story interesting.

An investor and a competitor can live inside the same company. Amazon has money in Anthropic, but it also competes in the same market with its own AI models and its cloud AI business. It can watch Anthropic’s growth through its equity stake with one hand while, with the other, preferring that Anthropic not get too far ahead. That’s where the logic of the theory begins.

The Real Threat Isn’t Money. It’s Power.

Here’s what separates this from a garden-variety corporate spat. The weapon isn’t market competition. It’s government power.

Big Tech fights have mostly played out inside the market. Cut prices, ship a better product, poach the best engineers. Exhausting for everyone involved, sure, but consumers at least got better products out of the brawl.

Now imagine an investor mobilizing its influence to move regulators, and using that regulation to push a rival’s model out of the market. The game changes completely. It’s no longer about who builds the better product. It’s about who sits closer to power. What comes back to consumers isn’t smarter AI. It’s a narrower menu.

Why the Open Source Camp Revolted

It’s no accident that the “open source AI must win” movement detonated alongside this episode.

Closed models are, by design, held in a small number of controllable hands. The fact that a government can order someone’s access cut off is, paradoxically, proof that the model sits at a controllable choke point. Flip one company’s API switch and it’s over.

Open models are different. Once they’re out, no one can switch them all off at once. So when developers watching this unfold start shouting that the answer is a model they can hold themselves, it follows naturally. It’s a refusal of the very structure that lets power reach for the off switch.

So What Should We Actually Watch?

Again, to be clear: it is not an established fact that Amazon moved the government. This is circumstance and suspicion. But the speed with which that suspicion became plausible tells you something real about the power structure of AI right now.

A world where a giant investor holds both the equity and the fate of a rival, and where distance from the regulator becomes a competitive advantage. If that’s the AI ecosystem we’re heading toward, the truly frightening question may not be which model is smarter.

So here’s what lingers. The next time the government flips the switch on some AI model, will we be able to tell whether the decision was made for safety or for someone’s bottom line? The moment we can’t answer that is the most dangerous moment of all.