Amazon’s CEO was among executives who privately raised concerns about Anthropic’s advanced AI

Amazon CEO Andy Jassy was among a group of technology executives who privately raised concerns with senior Trump administration officials this week about the security implications of Anthropic's most advanced artificial intelligence systems, a person familiar with the discussions told Reuters.

The intervention came as the White House moved to impose one of the most sweeping restrictions yet on frontier AI technology. On Friday, the Trump administration directed Anthropic to block all foreign nationals, whether located inside or outside the United States, from accessing its newest models, Claude Fable 5 and Mythos 5, citing national security concerns.

The decision marks a dramatic escalation in Washington's efforts to control the dissemination of cutting-edge AI capabilities and underscores growing anxiety within both government and industry over the potential misuse of increasingly powerful models.

Anthropic responded by announcing that it would disable access to the models globally rather than attempt to selectively enforce the restrictions. In a blog post, the company said U.S. officials believe a method exists to bypass, or "jailbreak," safeguards designed to prevent Fable 5 from being used to identify software vulnerabilities.

The restrictions were imposed through export-control mechanisms overseen by the U.S. Commerce Department's Bureau of Industry and Security (BIS), placing advanced AI models alongside sensitive technologies that Washington sees as strategic national assets.

Anthropic recently unveiled Claude Fable 5, the public version of its powerful Mythos system, which had previously generated alarm within cybersecurity and government circles after demonstrating an exceptional ability to discover software flaws. Anthropic has also confidentially filed for a U.S. initial public offering, placing its business strategy and relationship with regulators under intense scrutiny.

The Trump administration officials have become increasingly concerned that advanced AI systems capable of identifying vulnerabilities, accelerating scientific research, or assisting software development could also be exploited by hostile governments, cybercriminals, or military organizations.

Frontier AI models are now seen as dual-use technologies whose capabilities may have implications for cybersecurity, intelligence gathering, and military competition.

The restrictions form part of a broader strategy of the United States to treat advanced AI models similarly to semiconductors and other important technologies that have been subject to export controls in recent years.

However, Andy Jassy's reported involvement is notable because Amazon is not only one of the world's largest cloud computing companies but also one of Anthropic's largest financial backers. Amazon has invested billions of dollars in Anthropic and integrated Claude models across its cloud and AI offerings. Concerns raised by Jassy, therefore, signal that even some of Anthropic's closest partners may share government worries about the security implications of capable AI systems.

Washington's approach, however, has been criticized. Some experts who support export controls on advanced AI technologies also believe that the latest order is unusually broad and could undermine research collaboration with allies.

Jimmy Goodrich, a senior fellow at the University of California's Institute for Global Conflict and Cooperation, criticized the decision.

"This was not well thought-out," Goodrich said. "It even bans Canadians and Brits employed at Anthropic from doing research and development."

His criticism reflects a growing debate over whether AI restrictions should target geopolitical rivals such as China, Russia and Iran or whether they should apply universally regardless of nationality. Some experts have argued that restricting access for citizens of allied nations could weaken collaborative research networks that have traditionally been central to technological innovation.

The latest restrictions arrive after months of tension between Anthropic and the Trump administration. The company has previously challenged government efforts to limit its operations, including legal disputes surrounding Pentagon-related restrictions on the use of its technology.

Recent court filings revealed that federal agencies had sought to curtail the deployment of Anthropic products after disagreements over military applications of its AI systems and safeguards related to autonomous weapons and surveillance. There had been signs that those disputes were easing in some parts of the government. Friday's export-control action, however, indicates security concerns surrounding Anthropic's most advanced models remain unresolved.

The implications of the decision for the global AI race stand tall. Industry experts note that the decision could have consequences far beyond Anthropic as the restrictions may reinforce a broader trend toward the fragmentation of AI development along geopolitical lines, with governments increasingly seeking to control who can access the most powerful systems.

For countries attempting to build domestic AI capabilities, access to frontier American models has become an important source of knowledge and benchmarking. Limiting that access could make it more difficult for foreign researchers and companies to keep pace with leading U.S. developers.