Anthropic has suspended all foreign access to its most advanced artificial intelligence (AI) models — Fable 5 and Mythos 5 — after the US government issued an export control order citing national security concerns.
Export controls are typically used to restrict the supply of physical goods to other countries, such as advanced US graphics chips to China.
But the latest order goes beyond physical goods to dictate how software can be launched and distributed. This marks a new chapter in the complicated relationship between the US government and Anthropic and, more importantly, raises questions about the limits of AI if a government order simply cut off access to it.
Why the export control order?
Anthropic said that it received the US government’s order early on June 13, but the communication did not provide specific details of the administration’s national security concerns. According to the company, the government possibly believes that it has become aware of a method of bypassing, or “jailbreaking” Fable 5.
In a blog post, the company said that the order would block all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.
“We reviewed a demonstration of this specific technique being used to identify a small number of previously known, minor vulnerabilities. These vulnerabilities all appear relatively simple, and we have found that other publicly-available models are able to discover them as well without requiring a bypass,” Anthropic said.
But, it cautioned that the US government had only given Anthropic “verbal evidence” of a “potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws”.
Story continues below this ad
“Our understanding is that one potential jailbreak was shared with the government. We have reviewed a report that we believe is the basis of the government’s directive and validated that the level of capability displayed there is widely available from other models (including OpenAI’s GPT-5.5), and is used every day by the defenders who keep systems safe,” the company said in its blog post.
In March, after talks over how Anthropic’s AI could be used in military operations, the Pentagon had deemed Anthropic a supply chain risk, potentially limiting its use by federal agencies. This came after disagreements over how the military could use Claude, with Anthropic insisting on safeguards preventing the AI from being used for mass domestic surveillance or for developing fully autonomous weapons systems.
The order comes as Anthropic is preparing for public listing in the US, after being recently valued $965 billion.
What are Anthropic’s Fable 5 and Mythos 5?
Fable 5 and Mythos 5, launched earlier this week, are Anthropic’s most advanced publicly available AI models. They are built on the underlying “Mythos-class” architecture – the company’s other AI model named Mythos Preview which it has not released publicly over fears of misuse.
Story continues below this ad
Anthropic had previously claimed that Mythos Preview was able to find severe vulnerabilities in “every major operating system and web browser”, including one that had gone undetected for nearly three decades — spooking regulators around the world for its potential impact on critical infrastructure.
The company had started giving exclusive access to a select few organisations under its Project Glasswing. In India, Anthropic was discussing sharing Mythos Preview with a handful of entities, but was yet to share the model.
Since the company was not able to publicly release Mythos Preview, it announced earlier this week that it would launch Fable 5 and Mythos 5 instead, which are both based on the same underlying model, but have additional safeguards baked in.
In particular, requests related to sensitive cybersecurity or biological domains are filtered. When the system detects a potentially high-risk query, it automatically routes the user to a less capable model rather than allowing the request to be handled by the full Mythos-level system.
Story continues below this ad
“We’ve therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8. To release the model both safely and quickly, we’ve tuned these safeguards conservatively—they’ll sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions. With more capable models arriving in the coming months, we’re working to improve our safeguards and reduce false positives as quickly as we can,” Anthropic said in a blog post earlier.
The company also said that for a small group of cyber experts and infrastructure providers, it was launching Claude Mythos 5. It’s the same underlying model as Fable 5, but with the safeguards lifted in some areas.