As the U.S. government recently barred foreigners from using Anthropic’s latest artificial intelligence (AI) models Mythos and Fable, debate is growing over whether Anthropic’s long-running emphasis on AI risks has been used as justification for regulation.

IT outlet Ars Technica said on June 22 that the Financial Times reported an analysis of external communications by Anthropic and rival OpenAI this year found Anthropic far more actively stressed AI risks and the need for regulation.

The analysis covered official statements, social media posts and contributed articles by Anthropic and Chief Executive Dario Amodei. The Financial Times said Anthropic used terms related to risk, regulation and safety an average of 5 times per 1,000 words. The figure for OpenAI and CEO Sam Altman was 0.6.

Specifically, Anthropic used the term “risk” 336 times this year. It mentioned “safeguard” 121 times and “vulnerability” 128 times. OpenAI was at 30, 33 and 10, respectively.

In politics and the industry, some interpret the messaging strategy as linked to the U.S. government’s recent regulatory decision. The U.S. government last week barred foreigners from using Anthropic’s latest AI models Mythos and Fable. The move was reported to have been pursued on the grounds of national security and cybersecurity risks.

Mythos, at the center of the controversy, has been introduced as a high-performance model capable of detecting major cybersecurity vulnerabilities. Anthropic initially allowed access only on a limited basis to some U.S. institutions for safety reasons, and it was reported to have discussed launch plans with government officials until its formal release.

Critics say the process showed that “overemphasising risks ended up providing grounds for regulation.” Meta’s former chief AI scientist Yann LeCun criticised Amodei’s AI risk arguments on social media as “ridiculous fearmongering” and argued that he was “reaping what he sowed.”

Altman also mocked it in a podcast interview in April, mentioning Mythos and saying, “It’s quite a surprising marketing strategy to say we built a bomb and now we’re going to drop it on your head.”

Anthropic has been viewed as a leading “safety-first” voice in the AI industry. In a long blog post released shortly before the export ban, Amodei argued regulators were moving too slowly and warned AI could pose real risks to the financial system, critical infrastructure and national security.

But some point out that such warnings were used as logic to justify government regulation. David Sacks, a former AI czar who oversaw U.S. government AI policy, argued that “trusted partners suggested ways to bypass safeguards, but Anthropic downplayed it.” He said the government ultimately had no choice but to impose the ban “reluctantly.”

The move came as tensions between the U.S. government and Anthropic intensified. The two sides have publicly clashed over issues including the national security use of AI technology and the application of surveillance technology, and the U.S. Department of Defense in February designated Anthropic a supply-chain risk company on national security grounds. The two sides are now engaged in a legal dispute over the legality of the designation.

In the industry, some view the incident as not simply a problem limited to one company, but a first case showing the standards the U.S. government may use to control advanced AI models going forward. Questions are also being raised about policy consistency as the United States pushes to expand exports of advanced AI chips while imposing strong access restrictions on the most advanced AI models.

Experts expect the move to have no small impact on future talks on AI technology controls and regulation among the United States, Europe and key allies.