Why Anthropic’s Feud With the Government May Be a Net Win for Its Business

Anthropic’s latest clash with the U.S. government over access to its most advanced AI models may look like a setback, but fresh spending data suggests it is accelerating adoption among businesses rather than slowing it. In May, enterprise customers on Ramp’s platform spent more on Anthropic’s models than on OpenAI’s for the first time, a milestone that coincides with regulatory scrutiny, a record funding round, and the first-ever profitable quarter. The timing raises a counterintuitive question: when a company is branded a supply-chain risk and ordered to pull ultra-powerful models from the market, do customers flee—or flock?

A Government Ban That Looks Like a Product Endorsement

Last week, the White House invoked an export-control directive to bar non-U.S. individuals, including Anthropic employees, from accessing the company’s two newest models: Mythos 5, a restricted release, and Fable 5, a public-facing version released just three days earlier. The stated concern was that Fable 5’s guardrails could be bypassed by hackers, allowing access to capabilities comparable to Mythos 5. In response, Anthropic removed both models from availability rather than risk violating the order. The move effectively took its most capable offerings off the commercial market overnight.

The episode follows a March designation by the Department of Defense labeling Anthropic a supply-chain risk, a status usually intended to discourage government contractors from doing business with the company. Yet within weeks of that designation, Anthropic recorded its strongest month ever in business adoption, according to Ramp’s lead economist Ara Kharazian. “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,” Kharazian noted. The apparent paradox—government warnings fueling demand—points to a broader trend in enterprise AI: customers increasingly equate regulatory scrutiny with technical superiority and reliability.

Enterprise Spending Swings Toward Anthropic

Ramp’s data, drawn from more than 70,000 businesses using its corporate card and expense platform, shows that in May enterprise spending on Anthropic’s models surpassed spending on OpenAI’s for the first time. While the data does not break down exact dollar amounts or model-level breakdowns, the shift in relative spending is statistically meaningful across a large sample of organizations. The trend suggests that businesses are not just tolerating Anthropic’s controversies but actively choosing its models over alternatives at the point of purchase.

One plausible explanation is that the same qualities that make regulators uneasy—advanced capabilities, limited access, and stringent safety controls—are precisely what corporate buyers now seek. In industries such as financial services, healthcare, and cybersecurity, organizations are increasingly prioritizing models that can handle complex reasoning, detect software vulnerabilities, or automate compliance workflows. Anthropic’s emphasis on constitutional AI and its refusal to support mass surveillance or autonomous weapons have also resonated with enterprises that value ethical alignment. As one CIO put it in an internal memo reviewed by Ramp, “If the government is worried, we should be paying attention. That’s the model we need.”

Regulatory Scrutiny as a Market Signal

The government’s actions against Anthropic are not isolated. They follow a pattern in which agencies both warn about and, in some cases, attempt to restrict access to advanced AI systems. Yet the data shows that such scrutiny can function as a de facto endorsement. When a model is described as so powerful that it must be restricted, buyers interpret that as proof of capability. This is especially true in sectors where AI is used for high-stakes decision-making, such as threat detection, code analysis, and regulatory compliance.

The timing of the spending shift is also significant. It coincides with Anthropic’s announcement of a $65 billion valuation at the end of May, surpassing OpenAI’s reported valuation, and the filing of confidential IPO paperwork in early June. While valuation and IPO prospects are distinct from product adoption, they reflect investor confidence that is likely reinforced by strong enterprise traction. The fact that Anthropic achieved this milestone during a period of heightened regulatory pressure suggests that the market views regulatory scrutiny not as a liability but as a signal of differentiation.

The Mythos Effect: Why “Too Dangerous” Is Now a Selling Point

Mythos 5 is described by Anthropic as capable of finding subtle security flaws in large codebases—capabilities that the company itself has marketed as dangerous if misused. The model’s release notes explicitly caution that it can expose vulnerabilities that could be exploited by attackers if not handled with extreme care. This dual-use framing creates a paradox: the more dangerous the model, the more valuable it becomes to organizations that need to identify and fix vulnerabilities before they are weaponized.

Fable 5, released publicly three days before the government order, inherited some of that risk profile. Although it included guardrails intended to prevent misuse, reports indicated those controls could be bypassed by skilled attackers. Anthropic’s decision to withdraw both models underscores the tension between innovation and safety, but it also amplifies the narrative that Anthropic is at the frontier of AI capability. For enterprise customers, that frontier is where cutting-edge performance meets real-world utility.

The result is a form of inverse marketing: instead of claiming “our model is safe,” Anthropic’s models are framed as powerful enough to require extraordinary safeguards. In a market where buyers are overwhelmed by competing claims of safety and performance, this framing can cut through the noise. It signals not just capability, but a commitment to responsible innovation—even when that innovation is disruptive.

Financial Resilience in the Face of Restrictions

Anthropic’s ability to weather the withdrawal of its flagship models speaks to its financial resilience. The company reported its first-ever profitable quarter just before the regulatory actions, and its $65 billion valuation places it among the most valuable AI labs in the world. While the Ramp data does not quantify the revenue impact of removing Mythos 5 and Fable 5 from the market, the fact that enterprise spending continued to rise despite the loss of these products suggests that customers are pivoting to Anthropic’s existing suite—such as Claude 3.5 and earlier models—without a measurable drop in adoption.

This resilience is partly structural. Anthropic’s enterprise contracts are typically multi-year agreements with defined usage tiers, which can smooth out short-term disruptions. Moreover, the company’s focus on high-value, high-complexity use cases—such as software security audits, compliance automation, and risk modeling—means that customers are less likely to switch to cheaper, less capable alternatives when their core needs remain unmet.

What Happens Next: Regulatory Clarity and Product Evolution

The immediate question is whether the government will clarify or expand the export-control directive, potentially affecting not just access but also the development of future models. Anthropic has not indicated whether it will challenge the order or seek exemptions, but the episode highlights the need for clearer rules governing AI exports and dual-use technologies. Without such clarity, companies may face repeated disruptions as they navigate overlapping national security and commercial priorities.

For customers, the key takeaway is to expect continued volatility in access to the most advanced models. Organizations should plan for contingencies: diversify across providers, maintain fallback models, and ensure their workflows can adapt if a preferred model becomes unavailable. At the same time, they should evaluate Anthropic’s existing offerings for capabilities that meet their needs without relying solely on the latest restricted releases.

Anthropic, for its part, is likely to accelerate development of models that balance power with safety, while refining guardrail mechanisms to withstand bypass attempts. The company may also explore partnerships with cloud providers or governments to create compliant access pathways for restricted models, allowing it to re-enter the market without violating export rules.

Practical Takeaways for Enterprise Buyers

• Treat regulatory scrutiny as a signal, not a warning. If a model is restricted or labeled risky, investigate whether its capabilities align with your organization’s needs. The restriction may be a proxy for capability.
• Diversify your AI stack. Even if one provider gains traction, maintain relationships with at least two other model providers to mitigate supply-chain risks.
• Focus on use-case fit over hype. Evaluate models based on your specific workloads—security auditing, compliance, or code analysis—rather than chasing the latest release.
• Plan for access volatility. Assume that access to cutting-edge models may change rapidly due to regulatory or safety concerns. Build modular architectures that allow quick swaps.
• Engage with compliance teams early. When adopting restricted models, involve legal and security teams to ensure alignment with export controls and internal policies.

The Bigger Picture: AI Governance as a Competitive Differentiator

The Anthropic episode illustrates a broader shift in the AI market: governance is no longer a compliance checkbox but a competitive lever. Companies that can demonstrate robust safety frameworks, ethical alignment, and transparent risk management are winning enterprise trust—and budgets. Conversely, providers that prioritize speed over safety may find themselves at a disadvantage as buyers increasingly demand accountability.

This dynamic is likely to intensify as governments worldwide draft AI regulations. Organizations that align their AI strategies with emerging governance norms—such as the EU AI Act or U.S. export controls—will be better positioned to navigate future restrictions. For Anthropic, the current friction may ultimately reinforce its brand as a responsible innovator, provided it can translate governance into trust and reliability.

In the end, Anthropic’s feud with the government may prove to be a net win—not because the company avoided controversy, but because it embraced the role of the responsible disruptor. For enterprise buyers, that role is increasingly synonymous with credibility. The lesson is clear: in AI, being too dangerous to ignore is fast becoming the best way to be taken seriously.