Anthropic suspends access to Fable 5 after a US government directive

What happened

On June 12, 2026, at 5:21 PM ET, the US government issued an export-control directive requiring the suspension of access to Fable 5 and Mythos 5. Anthropic cut access for all users, globally — including its own employees who are foreign nationals. The measure targets only these two models; all other Anthropic models continue to operate normally.

The stated reason: a narrow "jailbreak"

The directive cites national-security concerns and the discovery of a possible "jailbreak" method (a way around the safeguards). According to Anthropic, the evidence consists of a "narrow, non-universal" jailbreak that involved "asking the model to read a specific codebase and fix any software flaws."

The company argues that this very capability "is used every day by the defenders who keep systems safe" and is "widely available from other models (including OpenAI's GPT-5.5)" — meaning it is not something specific or exclusive to Fable 5.

Anthropic's position

Anthropic reaffirmed its "defense in depth" strategy and stressed that "perfect jailbreak resistance is not currently possible for any model provider." The company warns that applying this standard "would essentially halt all new model deployments for all frontier model providers."

The key line from the statement: "We believe this is a misunderstanding and are working to restore access as soon as possible."

Update June 13: how Amazon triggered the directive

Reporting from June 13 by Axios and the Wall Street Journal revealed that Amazon researchers were the source of the government's report. According to people familiar with the matter cited by the WSJ, Amazon CEO Andy Jassy personally informed Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and National Cyber Director Sean Cairncross that Amazon researchers had used crafted prompts on Fable 5 to extract information about software vulnerabilities.

Katie Moussouris, CEO of cybersecurity firm Luta Security — who reviewed the Amazon findings on Anthropic's behalf — publicly rejected the government's characterization. She posted on BlueSky that the technique is in fact "Defense Oriented Prompting (DOP)", a methodology security practitioners use to systematically probe AI systems for vulnerabilities so those weaknesses can be remediated. Moussouris called the government's response "a complete overreaction," noting the information obtained is more useful to defenders than attackers: "It's not a jailbreak. It was Defense Oriented Prompting (DOP) — capabilities defenders need."

Access to Fable 5 and Mythos 5 remains suspended as of this update (June 14, 2026). Prediction markets on Kalshi priced a 68% probability of Fable 5 returning before July 1.

What it means for business

For most companies: no impact. Day-to-day automations — lead qualification, support, reporting, follow-up — run on Opus 4.8, which is unaffected by the directive. If your workflows didn't explicitly depend on Fable 5, nothing changes.

The underlying lesson: don't tie yourself to a single model. This episode shows exactly why an architecture that can switch between models matters. A well-built automation picks the right model per task and can fall back gracefully to an alternative if one becomes unavailable — whether for reasons of price, capability, or, as here, regulation. Dependence on a single frontier model is an operational risk, not just a cost decision.

If you were using Fable 5 for heavy projects (large code migrations, complex analysis), switch back to Opus 4.8 for continuity and watch Anthropic's announcements — the company says it is working to restore access.

Sources: ↗ Official Anthropic statement · ↗ Axios — How Amazon and the White House ended Anthropic's Fable · ↗ Fortune — "It's not a jailbreak" (Katie Moussouris, Luta Security)