Too Dangerous to Sell, Yet Valuable for Espionage: US Government Shuts Down Anthropic’s Leading AI Systems

Washington has abruptly deactivated two of the most advanced artificial intelligence systems globally, citing national security concerns. This decision has raised questions about the complexities and contradictions surrounding the use of AI technology.

At the same time that one branch of the U.S. government deemed the AI system Mythos too hazardous for international use, another branch was reportedly employing it to infiltrate the computer networks of adversaries. This duality highlights a significant conflict between the U.S. government and Anthropic, a company recognized for its stringent safety protocols in AI development.

The Trigger for Shutdown

On a Friday evening, Anthropic, the San Francisco-based firm known for the Claude chatbot, received a directive from the U.S. Commerce Department. By the end of that night, the company had shut down its two latest and most powerful systems, Fable 5 and Mythos 5, not just in the U.S. but worldwide. The order did not explicitly instruct Anthropic to deactivate its systems. Instead, it stated that the company could no longer provide access to any “foreign national,” which includes anyone who is not a U.S. citizen, regardless of their location. Faced with such a sweeping regulation, Anthropic concluded that the only way to comply was to turn off the systems for everyone, while older models like Opus 4.8 continued to operate.

Export Controls as a Tool

The government’s choice of action is noteworthy, as there is no specific U.S. law governing AI. Instead, the Commerce Department utilized existing export controls, a regulatory framework designed to prevent sensitive military technology from falling into the wrong hands. This same framework has been employed to restrict advanced computer chips from reaching China. Under these regulations, even demonstrating sensitive technology to a foreigner on U.S. soil is considered an export. Thus, an order aimed at foreign access effectively became a global shutdown, resulting in a commercial product being rendered inactive over a weekend.

Concerns Over AI Vulnerabilities

The official concern driving this decision is the possibility of a “jailbreak,” a term used to describe methods that bypass an AI’s safety protocols. According to reports, an administration official indicated that the action was prompted by a claim from a competing firm that it had discovered a way to circumvent Mythos’s safeguards. The systems would remain inactive until the government could bolster its cyber defenses, potentially taking weeks. However, Anthropic provided a different perspective, stating that it analyzed the alleged technique and found only minor, already known vulnerabilities—issues that many AI systems, including OpenAI’s GPT-5.5, could reveal without any special manipulation. The company asserted that no one had found a comprehensive method to breach the model’s defenses and labeled the situation as a misunderstanding. Anthropic cautioned that if a single minor flaw could justify the recall of a widely used product, no AI company could confidently release new technologies.

Lack of Transparency

The lack of clarity surrounding the government’s actions is significant. Anthropic noted that the letter it received did not specify the exact threat, and the government has not disclosed its evidence to the public. Anthropic has a vested interest in downplaying the threat, while the administration may have its reasons for amplifying it. Both narratives cannot be entirely accurate, and currently, neither can be substantiated.

Historical Context

The historical context of this situation is crucial. The very capabilities deemed too dangerous for export are the same capabilities that the U.S. government sought to control earlier this year. In July 2025, Anthropic entered into a two-year contract with the Pentagon valued at up to $200 million, becoming the first AI lab to operate its models on classified military networks. The relationship soured when Anthropic refused to abandon two stipulations: that its AI would not be used for mass surveillance of American citizens or to control fully autonomous weapons. The Pentagon insisted on access for “all lawful purposes,” leading to a breakdown in negotiations.

In late February, President Trump ordered federal agencies to cease using Anthropic’s tools, with Defense Secretary Pete Hegseth labeling the company a “supply chain risk,” a designation typically reserved for foreign entities like Huawei. Anthropic subsequently filed a lawsuit, and in late March, a federal judge granted an injunction that partially blocked the blacklist, indicating that the company was likely to prove the government’s actions were unlawful retaliation for its position on AI safety. However, part of the designation remained in effect, and the legal battle continues.

Contradictory Positions

Adding to the complexity, even as the Pentagon distanced itself from Anthropic, reports emerged that the National Security Agency had embedded several of the company’s engineers to assist in operating Mythos, potentially for offensive actions against nations like China and Iran. While it remains unclear whether the model was actively used in any attacks, the situation illustrates a stark contradiction: the technology is deemed too risky for government procurement and international sale, yet it is valuable enough for covert operations.

A New Paradigm for AI

These developments reflect a broader shift in how the U.S. government perceives artificial intelligence—not merely as software but as a strategic asset akin to advanced weaponry. This redefinition suggests that AI is something to be hoarded, withheld from competitors, and deployed against adversaries.

For developers, students, and businesses outside the U.S., the immediate impact is stark. Many have lost access to two of the most advanced AI systems available, with no input on the decision and uncertainty about when or if access will be restored. For investors, this incident raises concerns about the political risks associated with American AI products, which can be rendered obsolete overnight by government decree.

Whether the recent order represents a legitimate security measure or is simply another chapter in an ongoing dispute remains unclear. However, the underlying reality is evident: the most powerful AI technologies being developed in the U.S. are now seen as both a valuable asset and a potential threat, leading to a precarious situation for the technology and its stakeholders.

As reported by www.timesnownews.com.

Explore the latest digital editions of FAME Delivered in the Magazine section: https://famedelivered.com/magazine/

Published on 2026-06-14 04:26:00 • By FAME Delivered News Desk