US government suspends access to Anthropic's Fable 5 and Mythos 5 models

The U.S. Office of the Director of National Intelligence (ODNI) issued a directive on June 13, 2026 ordering Anthropic to block access to its Fable 5 and Mythos 5 language models for all U.S. users [Anthropic News]. The order cites “compliance concerns” related to export controls and national‑security assessments.

Anthropic confirmed that API keys tied to U.S. IP addresses were disabled at 09:00 UTC, and that the company is working with the agency to define a path for reinstatement. The shutdown affects any developer, startup, or enterprise that integrated the models into products ranging from chat assistants to code generators [Anthropic News].

The immediate impact is twofold. First, teams that relied on Fable 5’s 175 billion‑parameter text‑generation capability and Mythos 5’s multimodal reasoning must pivot to alternative models, likely incurring integration delays and additional licensing costs. Second, the directive signals that the U.S. government will enforce export‑control rules on frontier AI systems, a stance already seen in recent restrictions on other vendors.

Compliance teams will need to audit their AI‑toolchains for any prohibited services, and legal counsel should review the “Export Administration Regulations” cited in the order. The move also raises the prospect of further bans if additional models are deemed high‑risk.

Stakeholders should audit usage logs, identify fallback models, and document remediation steps to avoid operational disruption.