Anthropic closed May with 41% of business AI subscription spending, edging past OpenAI at 39.5% for the first time, according to data from corporate spend platform Ramp. Anthropic's share rose 2.5 percentage points month-over-month while OpenAI's stayed essentially flat. The crossover landed in the same week Anthropic raised $65 billion at a $965 billion valuation, and weeks before it filed confidential paperwork for an IPO on the back of what was reportedly its first profitable quarter.
Then on Friday, the White House sent Anthropic a letter demanding it block non-Americans, including Anthropic's own employees, from accessing its two most advanced models: Mythos 5, released to a limited group in April, and Fable 5, the more restricted public version that shipped three days before the order. Anthropic responded by pulling both models from the market entirely. The administration cited an export control directive, but the precise trigger has not been disclosed.
Chatter in security circles points to hackers bypassing Fable 5's guardrails, which were designed to wall off the underlying Mythos capabilities. Anthropic itself had marketed Mythos as dangerous on the grounds that it is unusually effective at finding security flaws in software code, and restricted public access accordingly. The March supply-chain risk designation from the Trump administration followed Anthropic's refusal to allow government use of its models for mass surveillance of Americans and for fully autonomous weapons.
Key facts
- 01Anthropic's share of business AI subscriptions hit 41% in May, up 2.5 points, edging OpenAI's 39.5%.
- 02The company raised $65B at a $965B valuation at the end of May, then filed confidential IPO paperwork in June.
- 03The Trump administration ordered Anthropic to block non-Americans from Mythos 5 and Fable 5, forcing both models off the market.
- 04Ramp's data covers more than 70,000 businesses, with Claude Opus dominating the API spend it can identify.
Ramp's lead economist Ara Kharazian, who compiled the business-spending data, argues the latest confrontation will not dent Anthropic's enterprise momentum and could amplify it.
The pattern Kharazian is pointing to has receipts. In March, when the Department of Defense formally labeled Anthropic a supply-chain risk, business adoption of its models posted a record month on Ramp's platform. The same dynamic — government friction acting as a credibility signal for enterprise buyers — appears to be repeating with the Mythos and Fable 5 episode.
Whether Anthropic takes a measurable revenue hit from pulling Mythos 5 and Fable 5 is harder to assess. Ramp's data isn't granular enough to isolate spend on those specific models.
What the data does show is heavy and growing business use of Anthropic's Opus line. Ramp draws on transaction data from more than 70,000 businesses. In the roughly one-third of transactions where model details are visible, spending skews toward various Claude Opus versions, particularly recent ones. Opus is the model line that preceded Mythos and remains openly available; Anthropic released Opus 4.8 in late May.
The bulk of enterprise AI spending isn't subscriptions at all — it's API calls measured in tokens, used for tasks like coding. Claude Code has built a reputation as one of the stronger coding tools on the market, and it is the workload driving much of the Opus consumption Ramp can see. OpenAI still leads Anthropic by a wide margin in overall consumer usage, according to Sensor Tower data, but the enterprise picture is now reversed.
The IPO question is the open one. Public-market investors typically discount companies tangled in active disputes with the federal government, and Anthropic now has two such disputes in three months. A confidential filing gives the company room to wait out the news cycle before pricing, but the Mythos guardrail bypass — and the cybersecurity researchers publicly questioning Fable's defenses — will follow the prospectus into roadshow meetings.
The Ramp numbers reframe the standard narrative about regulatory friction. For a frontier model lab selling to Fortune 500 buyers, being named too dangerous to use is closer to a marketing asset than a liability. Anthropic's enterprise customers are paying for capability, and a government designation is a free third-party endorsement that the capability is real. The harder question is whether public-market investors will read the same signal the same way when Anthropic prices its IPO, or whether they'll see a company whose best products keep getting pulled off the shelf.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
