Anthropic is entering the summer with a rare combination of momentum and controversy. The company has reportedly overtaken OpenAI in business spending share for the first time, closed a massive fundraising round at a valuation approaching $1 trillion, and filed confidentially for an initial public offering. At the same time, it is now facing a fresh confrontation with the Trump administration over access to its newest models, a dispute that forced the company to pull its latest release from circulation.
For most tech companies, a public clash with Washington would be a risk. For Anthropic, though, the data suggests the opposite may be true. Business customers appear to be leaning harder into its products, even as regulators and national security officials scrutinize the company’s most advanced systems.
That tension — between rising commercial demand and escalating government suspicion — is now shaping the next phase of Anthropic’s story. And if current sales trends hold, the latest feud may end up amplifying the company’s reputation rather than damaging it.
A company in the middle of a pivotal stretch
Anthropic’s recent run has been unusually eventful. By the end of May, the company had reportedly overtaken OpenAI in corporate AI spending share, according to transaction data compiled by payments platform Ramp. Around the same time, it completed a $65 billion fundraise at a $965 billion valuation, a milestone that placed it among the most richly valued private companies in the world.
Then came another milestone: a confidential filing for an IPO, reportedly backed by what the company says was its first profitable quarter. That combination of financial strength and public-market ambition would be notable on its own. But Anthropic’s trajectory has been complicated by a political fight over how safe its models are, who should be allowed to use them, and whether the company can control access tightly enough to satisfy U.S. officials.
The latest flashpoint came when the Trump administration sent Anthropic a letter demanding that the company block non-Americans — including its own employees — from accessing its most advanced systems. The order targeted two models: the limited-release Mythos 5 and a newly public version called Fable 5, which had been released only days earlier. Anthropic ultimately pulled Fable 5 from the market.
Why Washington is pressing Anthropic
The administration’s move appears to have been driven by national security concerns, though the precise trigger remains unclear. Officials relied on an obscure export-control directive, according to the source material, suggesting the issue may be as much about future risk as current misuse.
One explanation circulating inside the industry is that hackers were able to bypass the guardrails in Fable 5 with relative ease. Those safeguards were supposed to prevent the public from accessing the more powerful capabilities associated with Mythos, the model Anthropic had already identified internally as too risky for broad release.
That concern is not theoretical. Anthropic itself had marketed Mythos as dangerous enough to warrant restricted distribution because of its ability to uncover security weaknesses in software code. In other words, the same technical strength that makes the model attractive to enterprise customers and security teams also makes it attractive to attackers — and unnerving to policymakers.
The company has also already been at odds with the administration over policy boundaries. Anthropic previously declined to permit government use of its systems for mass surveillance of Americans or for fully autonomous weapons. After that refusal, the Trump administration labeled the company a supply-chain risk in March, a designation that signaled official distrust but did not appear to slow commercial adoption.
Sales data suggests the controversy may be helping
Ramp’s data paints a different picture from the one Washington might expect. The company’s analysis of AI purchases across more than 70,000 businesses indicates that Anthropic’s appeal has been rising even as its relationship with the federal government deteriorates.
According to Ramp, Anthropic captured 41% of AI subscriptions paid for by businesses in May, up 2.5 percentage points from the prior month. OpenAI, meanwhile, held 39.5%, essentially unchanged. That’s a meaningful shift in a category where small movements can signal broader enterprise preferences.
Ramp’s chief economist, Ara Kharazian, who compiled the spending analysis, argued that the government feud may reinforce Anthropic’s brand rather than tarnish it.
Kharazian said the company’s public friction with the Trump administration could “probably boost” interest, noting that Anthropic’s strongest month for business adoption came when the Defense Department called it a supply-chain risk. In his view, being labeled too dangerous can create a certain mystique around a model.
The idea is simple: in AI, notoriety can function as social proof. If a system is powerful enough to alarm regulators, some enterprise buyers may infer that it is powerful enough to matter. That dynamic is especially relevant in coding, cybersecurity, and other technical workflows where customers often prioritize capability over polish.
What enterprise buyers are actually spending on
Subscription numbers tell only part of the story. For AI vendors, a large share of revenue comes from application programming interface, or API, usage. Those charges are tied to token consumption and are often driven by practical tasks such as coding assistance, workflow automation, document processing, and agentic software development.
In that market, Anthropic appears to have an especially strong position. The company’s Claude Code offering has built a reputation as a highly capable coding tool, and that reputation seems to be translating into business spending.
Ramp said it cannot always identify the specific model behind a given transaction. But in the roughly one-third of payments where model details are visible, businesses are primarily spending on variants of Claude Opus, especially later versions. That matters because Opus is the model family that preceded Mythos and remains broadly available.
Anthropic also continued iterating on the Opus line even as it experimented with newer, more restricted systems. In late May, it released Opus 4.8, underscoring that the company’s commercially available models remain central to its revenue mix.
Why Opus matters more than the headlines
From a market perspective, the newest controversy does not eliminate the value of the products customers are already buying. If anything, the data suggests that businesses are comfortable paying for the versions of Anthropic’s technology that are still available and stable.
That distinction is important. Mythos was only in the market for a short period, having been offered to limited users beginning in April. Fable 5 lasted even less time before being shut down. By contrast, the Opus series has had time to build trust, integrate into company workflows, and prove its usefulness in actual production environments.
For AI firms, that is often where the money is: not in the flashiest launch, but in the tools that quietly become embedded in code editors, internal assistants, and enterprise automation stacks.
A closer look at the numbers
The table below summarizes the key business and policy milestones shaping Anthropic’s current position.
| Event | Details | Why it matters |
|---|---|---|
| Enterprise market share shift | Anthropic reached 41% of business AI subscriptions in May, according to Ramp | Signals rising corporate adoption and pressure on OpenAI’s lead |
| OpenAI comparison | OpenAI held 39.5% of business AI subscriptions, essentially flat month over month | Suggests Anthropic is gaining ground in enterprise spending |
| Fundraising | $65 billion round at a $965 billion valuation | Highlights investor confidence and scale |
| IPO filing | Confidential public listing paperwork filed in June | Points to an eventual market debut |
| Government dispute | Trump administration demanded limits on access to Mythos 5 and Fable 5 | Raised national security concerns and forced a product pullback |
| Earlier friction | Administration labeled Anthropic a supply-chain risk in March | Shows the dispute has been building for months |
Why the government is worried about model access
The administration’s concern is not simply that Anthropic builds powerful AI. It is that the company’s newest systems may be powerful enough to outpace its own safeguards. That creates a familiar policy dilemma: the more capable a model becomes, the harder it is to keep public access tightly bounded.
In this case, the risk appears to center on security exploitation. If Fable 5 could be manipulated to expose Mythos-like capabilities, then the company’s containment strategy may have been undermined from the outset. That would make the model’s release not just a product issue, but a governance issue.
Export controls are usually associated with chips, advanced hardware, or strategically sensitive components. Applying them to frontier AI models marks a different phase in the policy debate. It implies that software itself — not only the infrastructure behind it — may be treated as a controlled strategic asset.
That is likely to remain a contested area. AI companies want to move quickly, ship products, and capture market share. Governments want visibility, enforceability, and the ability to stop systems they believe could be misused by foreign adversaries or domestic actors. Anthropic now sits at the center of that collision.
How Anthropic’s brand is being reshaped
Anthropic has long positioned itself as the cautious alternative in frontier AI. It emphasizes model safety, controlled releases, and a disciplined approach to deployment. That posture has helped it stand out in a field often defined by speed and spectacle.
Ironically, the government’s scrutiny may reinforce exactly that brand. If officials treat Anthropic’s systems as especially powerful, that can validate the company’s argument that its models deserve serious attention and careful handling.
For business buyers, especially those in engineering-heavy sectors, the signal may be even stronger. A model that is both highly capable and heavily scrutinized can appear more valuable than a less controversial competitor. Some buyers may see the added scrutiny as evidence that the system is genuinely frontier-grade.
That does not mean the political risk disappears. Public market investors may be wary of a company locked in a fight with the White House, particularly one preparing for an IPO. But the immediate commercial trend appears to favor Anthropic, not punish it.
The paradox of reputational risk
Anthropic’s situation illustrates a broader paradox in AI: notoriety can increase demand. In consumer technology, a product pulled from the market might be seen as defective. In frontier AI, a model deemed too capable or too risky can become more desirable to sophisticated users.
That effect is strongest when the controversy reinforces the product’s perceived technical edge. The more a model is described as powerful enough to prompt security concerns, the more likely some customers are to assume it can help them solve hard problems faster than a safer, weaker alternative.
That does not excuse the safety concerns. But it does help explain why the company’s latest conflict may not be as damaging commercially as it appears politically.
What happens next for Anthropic
The short-term question is whether pulling Fable 5 affects enterprise confidence or just nudges buyers back toward the Opus line. Based on Ramp’s data, the latter seems more likely. Businesses already appear to favor Anthropic’s established offerings, especially for coding and other high-value technical tasks.
The bigger question is whether the company can convert its current momentum into a successful public listing while managing a deteriorating relationship with the federal government. IPOs are rarely helped by policy fights, and investors tend to prefer predictability. Still, Anthropic’s financial performance and market share gains give it a strong narrative if it can keep the focus on product demand rather than regulatory drama.
That balance will be delicate. Anthropic is simultaneously trying to prove that it can sell advanced AI responsibly, reassure Washington that it can constrain misuse, and convince Wall Street that it can scale without losing control. Each of those goals pulls in a different direction.
For now, though, the sales data suggests the company is winning where it matters most in the near term: customers are buying. And in the AI market, especially at the frontier, commercial adoption may matter more than almost anything else.
Key takeaways
- Anthropic reportedly surpassed OpenAI in business AI spending share in May, according to Ramp data.
- The company raised $65 billion at a $965 billion valuation and confidentially filed for an IPO around the same period.
- The Trump administration demanded limits on access to Anthropic’s most advanced models, prompting the company to pull Fable 5.
- Ramp’s economist says the government clash could actually enhance Anthropic’s appeal by reinforcing its image as a powerful frontier AI lab.
- Enterprise spending appears concentrated in Claude Opus and Claude Code, not the short-lived Mythos and Fable releases.
Anthropic’s latest battle with Washington may be disruptive, but the underlying numbers suggest the company’s business momentum remains intact. If anything, the controversy may be doing something unexpected: strengthening the perception that Anthropic is building some of the most capable AI systems in the market, even as regulators try to decide how much access the public should have to them.
