Breaking

Anthropic Disables Fable 5 and Mythos 5 Worldwide After US Order

June 16, 2026 at 15:22 EDT

Policy & Regulation
Foundation Models
Security

After the US Commerce Department ordered access cut off for foreign nationals on national security grounds, Anthropic on June 12, 2026 immediately disabled its latest models "Claude Fable 5" and "Claude Mythos 5" for all users worldwide. In an official statement, the company said selective restriction based on nationality was technically impossible—API keys and similar mechanisms cannot determine a user's nationality in real time—forcing it to disable both models uniformly. Other models such as Opus 4.8 are unaffected.

June 12, 2026 · US Commerce Directive

Three Days Live, Then Gone: US Orders Anthropic to Pull Its Top Models

An export-control directive ordered Anthropic to suspend foreign-national access to Claude Fable 5 and Mythos 5. To ensure compliance, the company disabled both models for every user worldwide — just 72 hours after launch.

3 days

From June 9 launch to the recall directive

~72 hrs

Total time the models were available before global shutdown

1st

Reported use of 2018 Export Control Reform Act authorities for AI models

Benchmark — SWE-Bench Pro

Why it mattered: a clear lead on coding

80.3%

Fable / Mythos 5

69.2%

Opus 4.8

58.6%

GPT-5.5

FrontierCode (Diamond)

29.3% vs 13.4% Opus

Humanity's Last Exam

59.0% vs 49.8% Opus

GDPval-AA

1932 vs 1890 Opus

Priced at $10 / 1M input and $50 / 1M output tokens — less than half the cost of the prior Mythos Preview. Stripe migrated a 50M-line codebase in a single day.

The official rationale vs. Anthropic's position

US officials cited risks of diversion to military or intelligence users in countries of concern such as China and Russia — invoking a narrow "jailbreak" technique affecting limited cybersecurity scenarios.

Anthropic said the directive included no written specifics, noted similar capabilities exist in public models like GPT-5.5, and warned the broad recall could set a precedent that halts frontier deployments industry-wide.

Praised

State-of-the-art on coding benchmarks
Unlocked long-horizon agentic tasks
Major price cut vs. prior model

Concerns

Overly broad safeguards causing frequent fallbacks
High cost and token use at max effort
Recall may push demand toward low-cost Chinese open-source alternatives

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.