Few things in tech are as strange as watching a company get simultaneously banned by the US government and embraced by the business world with open arms. That’s exactly what’s happening to Anthropic right now.
On Friday, the Trump administration ordered the AI lab to bar non-Americans, including its own employees, from accessing its state-of-the-art models. We’re talking about Mythos 5, the limited-release powerhouse, and Fable 5, the public version that lasted all of three days before being pulled. The White House invoked some obscure export control directive, but the buzz around the industry is that hackers had already figured out how to sidestep Fable 5’s guardrails. That model’s so good at finding security flaws in code that Anthropic itself marketed it as too dangerous for wide release. You know things are intense when the company that built the thing warns people about using it.
But here’s where it gets really interesting.
Just a week earlier, Anthropic was riding high on a wave of momentum that would make any startup jealous. The company raised $65 billion at a $965 billion valuation at the end of May, besting OpenAI on both counts. It filed confidential paperwork for an IPO. And for the first time ever, it surpassed OpenAI in market share of business spending, according to data from Ramp.
Let that sink in. Anthropic beat OpenAI at their own game, and it happened in the same month the Trump administration was publicly trying to kneecap them.
The timeline is almost ironic in its timing. In March, the Trump administration declared Anthropic a supply-chain risk after the company famously refused to let the government use its models for mass surveillance of Americans or fully autonomous weapons. That felt like it should have been a death sentence for a company’s government contracts and, presumably, its appeal to cautious enterprise buyers.
Instead, something else happened. Business adoption didn’t just survive the controversy, it exploded.
Ramp’s data, drawn from more than 70,000 businesses using its platform, shows Anthropic’s share of AI subscriptions paid for by businesses jumped 2.5 percentage points in May to hit 41%. OpenAI sat at 39.5%, essentially flat from the prior month. The lead economist who compiled these numbers, Ara Kharazian, put it bluntly: “Anthropic’s best month on record, as far as business adoption, was the month that the Department of Defense labeled them a supply-chain risk. There’s a lot of aura that comes with your model specifically being named too dangerous to use.”
There’s something almost paradoxically appealing about being too controversial for the government. It signals power. It signals that what you’re building actually matters. Companies don’t want to bet on the safe, boringchoice when the controversial option is clearly winning.
The numbers back that sentiment up across the board. Beyond subscriptions, most business AI spending goes toward API calls covering token usage for things like coding. Anthropic’s Claude Code has built a serious reputation as a powerhouse coding tool. When Ramp can actually see which models businesses are using in their transactions, about a third of the time they show up clearly and businesses are overwhelmingly spending on various flavors of Claude Opus, particularly the later versions. Opus preceded Mythos and remains openly available, and it seems businesses aren’t exactly upset about having to stick with it.
Late May brought Opus 4.8, a fresh release keeping the line relevant even as the controversy around Mythos and Fable 5 swirls.
So what happens next? It’s hard to say. Pulling Mythos and Fable 5 off the market obviously costs Anthropic something, though exactly how much remains unclear given the limited time those models were available. The IPO path gets messier when you’re in an active fight with the federal government. Public market investors tend to be allergic to that kind of headline risk.
But the business fundamentals tell a different story. Companies are voting with their budgets, and right now they’re choosing Anthropic over OpenAI at rates that would have seemed impossible a year ago. The available models are more popular than ever, the valuation is staggering, and despite the government’s best efforts to paint them as a risk, enterprise customers see something else entirely.
Maybe being too powerful to trust is exactly the selling point enterprises didn’t know they were looking for.
