Anthropic market share rises as Mythos and Fable 5 are pulled

Anthropic pulled its limited-release Mythos 5 and the shortly available Fable 5 after the Trump administration sent a letter ordering a ban on non-Americans, including Anthropic employees, from accessing those models. Ramp’s spending data shows Anthropic’s share of AI subscriptions paid for by businesses rose 2.5 percentage points in May to 41%.

What happened to Mythos 5 and Fable 5?

The White House sent a letter demanding that Anthropic block non-American access to Mythos 5 and Fable 5, invoking an obscure export control directive, which effectively forced Anthropic to remove the models from the market. Mythos had been released to a limited set of users in April, and Fable 5 was released to the public three days before being shut down after a few days of availability.

The source of the ban is unclear in public filings, though chatter suggested hackers could bypass Fable 5’s guardrails that were intended to limit access to Mythos’ capabilities. Anthropic had previously described Mythos as powerful enough to find security flaws in software code, a reason the lab presented for restricting its public rollout.

How strong is Anthropic’s business adoption right now?

Ramp’s dataset, drawn from more than 70,000 businesses that use its platform, shows growing enterprise use of Anthropic’s models: Anthropic’s share of AI subscriptions paid by businesses increased 2.5 percentage points in May to 41%, while OpenAI held 39.5% and was essentially flat from the prior month. Ramp’s lead economist Ara Kharazian says the publicity may boost adoption: "If anything, it’ll probably boost them."

Ramp cannot always identify which model a given transaction targets; it can see model details in about one-third of transactions. When model-level details are visible, businesses are mostly spending on various flavors of Claude Opus, particularly later versions. Anthropic released Opus 4.8 in late May, and Opus remains openly available while Mythos had been restricted.

Beyond subscriptions, most company spending goes to API calls that bill tokens for tasks like coding. Anthropic’s Claude Code is noted in the data as having a strong reputation as a coding tool. Sensor Tower data cited alongside Ramp shows OpenAI still greatly leads Anthropic in overall consumer usage, even as Anthropic’s business share climbed.

Anthropic’s corporate milestones add context: the company raised $65 billion at a $965 billion valuation at the end of May, and in early June it filed confidential paperwork for an IPO, reportedly on the strength of its first-ever profitable quarter.

Why it matters

The White House action forced a high-profile model pull at a time when Anthropic was gaining enterprise momentum. That combination creates both regulatory and market risks: public-market investors historically shun companies entangled with government controversy, but the data suggests that explicit government concern can increase business demand. Ramp’s figures show the practical result: a leap to 41% of business-paid AI subscriptions, coupled with continued spending on Opus models used for coding and other tasks.

The episode also highlights the tension between model capability and safe deployment. Anthropic's restrictions on Mythos reflect internal judgments about risk, while the administration's supply-chain risk designation in March followed Anthropic’s refusals to let its models be used for mass surveillance and fully autonomous weapons.

What to watch

Watch whether Anthropic restores Mythos or Fable and how that affects both enterprise spending and investor appetite for the planned IPO. Also track Ramp’s next monthly snapshot for any reversal of the 2.5 percentage point move and Sensor Tower releases for shifts in consumer usage compared to enterprise trends.