Key Takeaways
- The U.S. Commerce Department ordered Anthropic to suspend access for all foreign nationals to its newest AI models, Fable 5 and Mythos 5, citing national‑security concerns.
- The directive also bars foreign‑national employees of Anthropic from using the models, marking an unusually broad restriction on who may employ a private company’s AI technology.
- This is the second Trump‑administration action against Anthropic; earlier the Pentagon labeled the firm an unacceptable supply‑chain risk for military and intelligence use.
- Anthropic said the order is a misunderstanding and pledged to restore access promptly, while noting that older models such as Opus 4.8 remain unaffected.
- The move has sparked debate among cybersecurity experts: some view Mythos 5 as a dangerous offensive tool, whereas others argue it can equally aid defense, and they question whether sweeping restrictions are warranted.
- Parallel developments include a White House executive order urging voluntary pre‑release oversight of AI systems and a similar, though less restrictive, limitation announced by rival OpenAI on its GPT‑Cyber 5.4 model.
Background on the Government Order
On Friday night Anthropic announced via social media that the U.S. government had directed the company to suspend access for all foreign nationals to its Fable 5 and Mythos 5 artificial‑intelligence models. The order, issued by the Commerce Department, was described as necessary to ensure compliance with national‑security concerns. Anthropic said it must “abruptly disable” the models for all customers to obey the directive, and the restriction also applies to Anthropic’s own foreign‑national employees. The company did not disclose how long the limits would remain in place, noting only that the decision came from a person familiar with the matter who spoke on condition of anonymity.
Why the Order Is Unusual
Limiting who can use a private firm’s AI technology based on nationality is rare in U.S. policy. While export controls on semiconductors and other hardware to countries such as China are common, restricting the users of a software product—especially to block citizens of allied nations like Canada or the United Kingdom—represents a far broader intervention. Analysts note that the move could hinder Anthropic’s research and development, as many of its engineers and collaborators hold foreign citizenship, potentially slowing progress on its cutting‑edge models.
Connection to Earlier Administration Actions
This restriction marks the second time the Trump administration has singled out Anthropic. In March, after tense discussions about how the company’s AI might be employed in military and intelligence contexts, the Pentagon declared Anthropic an unacceptable supply‑chain risk, which could limit its use by federal agencies. The latest order therefore continues a pattern of heightened scrutiny toward the San Francisco‑based firm, reflecting broader administration concerns about the dual‑use nature of advanced AI systems.
Anthropic’s Response and Impact on Operations
Anthropic declined to elaborate beyond its brief social‑media post, stating that it believes the government’s decision is a misunderstanding and that it is working to restore access as soon as possible. The company affirmed that its older models—including Opus 4.8 and other prior versions—were not affected by the order. Nevertheless, the suspension of Fable 5 and Mythos 5 could disrupt ongoing projects that rely on these systems for tasks ranging from advanced natural‑language processing to specialized cybersecurity analyses, potentially delaying product timelines and internal experimentation.
The Nature of Mythos 5 and Fable 5
Anthropic unveiled Claude Mythos in April, describing it as a powerful AI system that could be repurposed by hackers to breach computer networks. Because of this risk, the firm initially shared Mythos only with about 40 organizations that maintain critical infrastructure, allowing them to test and patch vulnerabilities before malicious actors could exploit them. Reactions from those early users were mixed: some viewed Mythos as a novel and alarming threat, while others considered it more evolutionary than revolutionary, arguing that the same capabilities could be used defensively as well as offensively.
Introduction of Fable 5 and Guardrails
Following the Mythos rollout, Anthropic released a “straitjacketed” version called Claude Fable 5, which incorporates additional guardrails intended to block responses in sensitive domains such as cybersecurity, biology, and other high‑risk topics. The company said that queries deemed risky would be rerouted to an older model, Opus 4.8. While these safeguards aim to prevent malicious hackers from weaponizing the AI, critics warn that they might also impede legitimate defensive work by cybersecurity professionals who rely on the model’s full capabilities to detect and mitigate threats.
White House Oversight Initiatives
In the wake of the Mythos announcement, the White House began contemplating new oversight mechanisms for AI systems. Last week President Trump signed an executive order asking companies to voluntarily submit new AI models for government review before public release. This shift marks a departure from the administration’s previously hands‑off stance on AI regulation and suggests a growing willingness to intervene in the development and deployment of powerful models, albeit on a voluntary basis for now.
Cybersecurity Community’s Divergent Views
Experts remain split on whether Anthropic’s decision to limit Mythos 5’s release was justified. Some argue that the model’s potential to facilitate network intrusions warrants strict controls, while others contend that the technology’s dual‑use nature means it can equally strengthen defenses. Notably, rival OpenAI announced a comparable limitation on its GPT‑Cyber 5.4 model but shared it with hundreds of experts initially, expanding to thousands later—a approach that contrasts with Anthropic’s more abrupt, blanket suspension for foreign nationals.
Reactions from Former Officials
Former U.S. cybersecurity and technology‑policy officials expressed surprise at the order, describing it as highly unusual and out of step with other Trump‑era moves, such as permitting the sale of advanced AI chips to China. Dean Ball, a senior fellow at the Foundation for American Innovation and a former Trump‑era AI adviser, called the decision “baffling” and said he had “no words” regarding the rationale. Their comments underscore the unease within the policy community about the precedent set by restricting access to AI tools based on users’ nationality rather than on the nature of the technology itself.
Broader Implications for the AI Industry
The episode highlights the growing tension between innovation and national security in the AI sector. As models become more capable, governments may increasingly seek to dictate who can develop, access, or deploy them. For companies like Anthropic and OpenAI, navigating these pressures will require balancing compliance with maintaining open research ecosystems that foster rapid advances. How the industry adapts to such regulatory interventions could shape the future trajectory of AI development, international collaboration, and the overall security landscape.

