# Government Halts Anthropic's Advanced AI Models Amid Dual-Use Fears Late last week, AI firm **Anthropic** was compelled to take its new **Claude Fable 5** and **Mythos 5** AI models offline following a United States government export-control directive. The order specifically bars "any foreign national" from using the services, sparking ongoing discussions between **Anthropic** and the White House. ## The Dual-Edged Sword of AI Capabilities Since its debut in April, **Anthropic** has both touted and warned about the advanced capabilities of its **Mythos** model. The company noted its potential for not only identifying software vulnerabilities to aid defenders but also for developing exploits that could be leveraged by bad actors. In a blog post announcing **Mythos 5** and **Claude Fable 5**, **Anthropic** explicitly stated, "A great deal of advanced usage of AI models is dual use: the same queries that are beneficial in the hands of cybersecurity professionals and biology researchers could be dangerous if available to malicious actors." Initially, a version dubbed **Mythos Preview** was released to a select consortium as part of **Project Glasswing**. **Mythos 5** followed a private release to this group, while **Claude Fable 5**, a **Mythos**-grade model, was made publicly available with specific safeguards to block responses related to biology and cybersecurity. ## National Security Concerns Trigger Restriction The Trump administration's move to restrict both models was predicated on the belief that **Fable 5**'s guardrails could be bypassed, granting full access to the more potent **Mythos 5** capabilities, thereby posing a national security risk. ## A Broader Trend, Not an Isolated Incident Cybersecurity experts argue that this institutional clash, while focused on **Anthropic** today, merely delays or masks a broader truth: similar AI capabilities are rapidly emerging across the industry. "It's myopic in the extreme to think that no other competitors to **Anthropic** will develop similar capabilities to **Mythos** or even that they have not already done so," states Tarah Wheeler, Chief Security Officer of TPO Group. "There are other companies hot on **Anthropic**'s heels who probably have the capabilities, too, and are holding them in reserve as they see how **Anthropic** is being treated in the current regulatory environment." **Anthropic** itself has echoed this sentiment. Logan Graham, the company's frontier red team lead, told WIRED in April, "The real message is that this is not about the model or **Anthropic**. We need to prepare now for a world where these capabilities are broadly available in 6, 12, 24 months." Echoing this trend, **OpenAI** also privately released a cybersecurity-focused model in mid-April and announced an expanded cybersecurity strategy. ## The Inevitability of AI Advancement Researchers highlight that even current AI offerings, with refined prompting, can be used for advanced vulnerability hunting and exploit development. This point was underscored in an open letter to the administration from a large group of cybersecurity leaders, who argued that the White House's export-control directive was misguided. "It's not one model; it's the general trend of technology," explains Bruce Schneier, a researcher at Harvard University and the University of Toronto. "Smaller, cheaper, open-source models, sometimes by themselves and sometimes in concert with each other, can match **Mythos**/**Fable**'s performance with more sophisticated prompting. And we should expect other models to match **Mythos**/**Fable**'s creativity and tenaciousness within months—slightly longer for open-source models." ## The Path Forward: Broader, Transparent Regulation Experts stress that governments globally need to focus on democratically developing broader and more transparent plans to contend with the inevitable advancements in AI capabilities across cybersecurity and other sensitive sectors. Chris Wysopal, cofounder of **Veracode**, summarizes the core policy question: "The question is whether a specific restriction meaningfully reduces that risk or whether it mainly slows down the people trying to make systems safer."