Anthropic Got Called Too Dangerous. Sales Climbed Anyway.

Being labeled a risk by the government was supposed to hurt Anthropic. The spending data says the opposite.

Anthropic just took its most powerful AI off the market after a federal export directive, and its business sales look set to grow because of it, not despite it. The reason is reputation: per Ramp data from over 70,000 companies, Anthropic’s share of business AI subscriptions rose 2.5 points in May to 41%, passing OpenAI for the first time.

Picture a product so capable the government tells the company to pull it from shelves, and the order doubles as a five-star review. That is roughly where Anthropic finds itself this month. The “too dangerous to use” sticker that should scare off buyers is instead acting like a status symbol, and the company’s numbers were already pointing up before the latest drama even started.

What Happened

On Friday, Washington renewed pressure on the AI lab, sending a letter demanding it block non-Americans, including its own employees, from accessing its top models: the limited-release Mythos 5 and the more guarded public version called Fable 5. The order effectively forced Anthropic to pull its flagship model from the market entirely.

The directive cited an obscure export-control rule, though the exact trigger remains unclear. The chatter points to hackers easily slipping past Fable 5’s guardrails, which were built to keep the public from reaching Mythos-level capabilities. That model is so sharp at finding security flaws in software that Anthropic itself flagged it as dangerous and restricted its release.

The Backstory

This is not the first standoff. Back in March, the company was declared a supply-chain risk after it refused to let the government use its models for mass surveillance and fully autonomous weapons. By most playbooks, a public fight with regulators is a sales killer.

Except it was not. Here is the line that should make every founder sit up, from Ramp lead economist Ara Kharazian, who compiled the data: “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.”

His read on the current feud? “If anything, it’ll probably boost them.”

The Numbers Behind the Run

Anthropic is having a month. It closed May by surpassing OpenAI in business spending share for the first time, raised $65 billion at a $965 billion valuation, then opened June by filing confidential IPO paperwork, reportedly on the back of its first-ever profitable quarter.

The Ramp data, drawn from more than 70,000 businesses, tells the operational story. Anthropic’s slice of business-paid AI subscriptions climbed 2.5 points to 41% in May. OpenAI sat at 39.5%, essentially flat. (OpenAI still leads heavily on consumer usage, per Sensor Tower.) Most company AI spend, though, goes to API calls for work like coding, where Anthropic’s Claude Code has a serious reputation. When Ramp can see which model is being used, in about a third of transactions, it is mostly various flavors of Claude Opus, the line that came before Mythos and is still openly for sale. The newest, Opus 4.8, shipped in late May.

The Business Model Angle

This is a lesson in earned scarcity. Anthropic’s entire brand is built on the claim that its models are powerful enough to be dangerous, and that it takes that danger seriously. When the government acts on exactly that premise, it does not contradict the pitch. It confirms it. The restriction becomes proof of capability.

For founders, the pattern is sharper than “all press is good press.” It only works when the controversy validates your core value proposition instead of undermining it. A safety scandal would have crushed a brand selling trust. A “this is too potent for open release” verdict, for a company selling frontier power, reads as a credential. The takeaway: know which story your brand is actually selling, because the same headline can be poison or rocket fuel depending on the promise underneath it. For more on how Anthropic built this position, see our breakdown of how Anthropic hit a $900B valuation and leapfrogged OpenAI.

The Risk

Now the cold water. Ramp’s data is not granular enough to show the actual revenue hit from pulling Mythos and Fable 5, so “aura” and “income statement” are not the same thing yet. Mythos only reached limited users in April, and Fable 5 was live for just a few days before shutdown, so the lost sales may be small for now. But a model you cannot sell generates no revenue regardless of how impressive its banishment looks.

The bigger overhang is the IPO. Public-market investors tend to flinch at companies tangled in government disputes, and Anthropic is walking toward its listing with an active one. Buzz can boost adoption of the models still on the shelf while quietly raising the risk premium on the stock. Reputation and valuation do not always move in the same direction.

Quick Questions

The Business Model Analyst Take

The counterintuitive headline is the whole lesson: a “too dangerous” label only sells if danger is your product. Anthropic spent years convincing the market its models were powerful enough to warrant caution, so when the government acted on that exact claim, the restriction read as proof rather than warning. Founders should clock the mechanism, not just the outcome. Controversy compounds your brand only when it confirms the promise you were already making. Pick a fight that validates your story, and the market may thank you for it. Pick the wrong one, and the same spotlight burns the house down. Just remember the gap Ramp could not measure: aura is not yet revenue, and a banned product still sells nothing.

Source: TechCrunch