Anthropic AI Models: Why Foreign Access to Fable 5 and Mythos 5 Was Restricted

Anthropic AI Models Latest News

• The US government has issued an export control order barring all foreign nationals — even those working inside Anthropic, a leading AI company — from accessing its two newest AI models, Fable 5 and Mythos 5. 
• These were launched recently and are Anthropic’s most advanced publicly available AI systems. 
• The order cites national security concerns, but its broad scope has raised fresh questions about how governments can control access to AI technology.

What Are Export Controls, and Why Is This Different

• Export controls are rules that stop certain goods or technologies from being sent to other countries. Normally, they apply to physical items — for example, the US restricts the sale of advanced computer chips to China.
• This new order is different. For the first time, it controls how an AI software product can be used and by whom — not a physical product. 
• This is being seen as a new and unusual use of export control law, applied to artificial intelligence for the first time in this manner.

‹ ›

Export Control Order Banning Access of AI Model

• The order blocks all foreign nationals from using Fable 5 and Mythos 5 — whether they are located inside the US or outside it. This includes Anthropic’s own foreign employees. 
• In simple terms, only American citizens or entities can access these two models for now.
• As per the Anthropic, the government did not give a clear, written explanation. Based on informal communication, the company believes the concern relates to a possible “jailbreak” — a method to bypass an AI model’s safety restrictions.
  • This technique works by asking the AI model to read a piece of software code and identify and fix flaws in it. 
• Anthropic has pushed back, stating that this capability is not unique to its models — other publicly available AI systems, including OpenAI’s GPT-5.5, can do the same thing. 
• In fact, this kind of capability is used daily by cybersecurity professionals to find and fix software weaknesses.

Background: Anthropic’s Troubled Relationship with the US Government

• This is not the first friction between Anthropic and US authorities. In March 2026, the Pentagon labelled Anthropic a “supply chain risk” — a designation that could limit how much US government agencies can use the company’s AI products.
• This labelling followed a disagreement over military use of Anthropic’s AI (Claude). 
  • Anthropic had insisted on certain safety conditions — specifically, that its AI should not be used for mass domestic surveillance or to help build fully autonomous weapons systems. 
  • The company has reportedly indicated it may legally challenge this “supply chain risk” label.
• This export control order comes at a sensitive time — Anthropic is preparing for a public stock market listing in the US, and was recently valued at $965 billion.

What Are Fable 5 and Mythos

• Anthropic has launched Fable 5 and Mythos 5, its most advanced publicly available AI models. 
• Both are built on the company’s new “Mythos-class” architecture, derived from the earlier Mythos Preview model, which was not publicly released due to concerns over potential misuse.

Why Was Mythos Preview Not Released

• Anthropic had previously claimed that Mythos Preview possessed the capability to identify severe vulnerabilities in major operating systems and web browsers, including some that reportedly remained undetected for years.
• The company feared that such powerful capabilities could be misused, particularly against critical infrastructure worldwide. As a result, it decided against a public release.
• Before the launch, Anthropic provided limited access to Mythos Preview through Project Glasswing. 
• In India, the company reportedly discussed sharing the model with a small number of organizations.

Key Difference Between Mythos Preview and Fable 5/Mythos 5

• Although Fable 5 and Mythos 5 are based on the same underlying technology as Mythos Preview, they include additional safety safeguards.
  • Requests related to sensitive cybersecurity issues or biological threats are filtered.
  • When the system detects a potentially high-risk query, it automatically redirects the user to a less capable AI model (Claude Opus 4.8) instead of allowing the full Mythos-level system to respond.
• The company acknowledges that the safeguards may occasionally block harmless requests.

Conclusion

• This episode highlights a growing global concern.
• As AI models become powerful enough to find security flaws in critical software, governments are beginning to treat advanced AI itself as a controlled, sensitive technology — similar to how they treat nuclear technology, advanced weapons, or top-end semiconductor chips.
• For India, this is a reminder that building indigenous AI capability is no longer optional but a matter of strategic necessity.

Source: IE | ToI