Anthropic's most powerful model yet launched, then went dark within days

It has been a whiplash week for anyone following Anthropic. On June 9, 2026, the company released Claude Fable 5, the first publicly available model in its new Mythos-class tier — a capability level the company positions above its long-running Opus line. Three days later, both Fable 5 and its bigger sibling Mythos 5 were switched off around the world.

What launched

Fable 5 arrived as Anthropic's most capable widely released model to date. The headline features: adaptive thinking that stays on by default, a context window of one million tokens (roughly the length of several long novels in one conversation), and support for very long outputs. The full Mythos 5 model was never opened to the general public; it remained limited to a small set of partners in a program Anthropic calls Project Glasswing.

Early testing chatter suggested a real step up in coding, research, and multi-step agent work — the kind of jump that usually resets the leaderboard for a few months.

Why it was pulled

On June 12, the US government issued an export-control directive citing national security concerns. According to reporting on the order, Commerce Secretary Howard Lutnick directed Anthropic to suspend access to Fable 5 and Mythos 5 for any foreign national, anywhere — including foreign nationals employed by Anthropic itself.

The trigger appears to have been a security report in which researchers demonstrated a way to bypass Fable 5's guardrails and get it to identify software vulnerabilities, raising fears the model could be misused as a cyber tool.

Because Anthropic has no practical way to verify the nationality of every user in real time, it chose the blunt option: it disabled both models globally while it works through the directive with regulators. As of this writing, there is no confirmed date for restored access.

The bigger picture

This is the first time a frontier AI model has been taken offline by a government order rather than a company decision. Whatever your view of the merits, it sets a precedent: model access can now change overnight for reasons that have nothing to do with pricing or product strategy.

Why it matters for small business

If your business has started leaning on one AI model for daily work — drafting, quoting, customer replies, coding — this week is a useful warning. Access can disappear with no notice, even from the most safety-focused lab in the industry. The practical fix is cheap: know which second-choice model you would switch to, keep your key prompts and workflows written down somewhere outside the chat tool, and avoid building anything mission-critical that only works with a single provider. You do not need a complicated backup plan — you just need one at all.