TL;DR
The US Department of Commerce lifted export controls on Anthropic’s Claude Fable 5 and Mythos 5 on June 30, with access restoration set to begin July 1 after 18 days offline. The shutdown showed that frontier AI access can be halted quickly by government order, while the security claims that triggered the move remain disputed.
On the evening of June 30, the US Department of Commerce lifted export controls on Anthropic’s Claude Fable 5 and Mythos 5, ending an 18-day government-ordered shutdown that had cut off access to two frontier AI models across major cloud platforms.
Anthropic launched Fable 5 on June 9, presenting it as the first publicly available model in its high-end Mythos class. Three days later, on June 12, Commerce sent CEO Dario Amodei a directive citing national-security authorities and ordering the company to suspend access for foreign nationals, including non-citizen employees, according to the source material.
The company was reportedly given about 90 minutes to comply. Because Anthropic could not filter users by nationality in real time, it took both models offline worldwide, affecting AWS Bedrock, Google Cloud, Microsoft Foundry and direct Claude API access within hours.
The path back came with new conditions. According to the source material, Anthropic agreed to proactively detect and address security risks, establish protocols for future model releases, report malicious activity found in models, and deploy a safeguard that Commerce’s CAISI tested against the reported jailbreak.
A frontier AI model went dark for 18 days. The kill-switch is real now.
Commerce lifted its export controls on Claude Fable 5 and Mythos 5, and access is being restored. But the reprieve isn’t the story — a state-of-the-art model was switched off by government order in an afternoon, and the deal to switch it back on wrote a new template for how frontier AI ships.
A frontier model now passes through a national-security gate before — and maybe after — release. It’s not isolated: OpenAI’s GPT-5.6 also went out to a small set of approved partners after a government request, and Mythos 5 returns first to government-approved customers. An August executive-order deadline for standardized AI-risk benchmarks points to formalizing the improvised process. The open question: does Washington now approve every frontier release?
The reprieve is real; the lasting change is the template. For builders the lesson is blunt and side-neutral: the firms that mapped their dependencies hot-swapped to alternatives (Claude Opus 4.8 among them); the rest went dark on 90 minutes’ notice. Model access is now a geopolitical variable, not a given. The rational answer isn’t loyalty to one lab or one government’s mood — it’s portability: multiple providers, tested fallbacks, and open-weight or self-hosted capacity you control. Don’t build as though access is permanent. It isn’t — now everyone’s seen the proof.
Model Access Became Conditional
The shutdown matters because frontier AI access was halted after release by a national-security order, not by a routine platform outage. For companies building products on these systems, the episode showed that model availability can become a policy risk as well as a technical or commercial one.
Businesses with tested fallbacks could move work to alternatives, including other Claude models, according to the source material. Those without multi-provider plans, portable prompts, or self-hosted capacity faced sudden disruption with little time to react.
Agentic AI Unleashed: A guide to designing, building, and deploying autonomous AI systems (English Edition)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
The Disputed Security Trigger
Why Commerce acted remains contested. According to Wall Street Journal reporting summarized in the source material, Amazon researchers claimed a sequence of prompts could jailbreak Fable 5 into producing information potentially useful for cyberattacks, and Amazon-White House discussions reportedly fed into the directive.
Anthropic disputed that account, describing the issue as a narrow potential vulnerability and warning that applying that bar broadly could halt all frontier-model deployment. The source material says independent analysts later argued the jailbreak reports had been overstated and questioned why rival models would not face the same treatment.
“a state-of-the-art model was switched off by government order”
— Thorsten Meyer AI, Reality Check AI Dispatch
Enterprise Security Frameworks: Access Control Models, Encryption Standards, and Audit Procedures for Regulated Industries (Mastery in Snowflake Cloud Data Management Series)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Jailbreak Evidence Still Unsettled
It is not yet clear how strong the technical evidence behind the directive was, how Commerce weighed competing security assessments, or whether similar issues exist in rival frontier models. The source material also leaves open whether Washington now expects pre-release approval or post-release review for every frontier model.
AI safety and jailbreak detection tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
August Benchmarks May Set Rules
Access restoration is set to begin on July 1, with Mythos 5 returning first to government-approved customers, according to the source material. An August executive-order deadline for standardized AI-risk benchmarks could turn the improvised review process into a clearer release gate.
AI-Powered Cybersecurity: AI Tools for Enterprise Security | AI for Network Security | AI Risk Management | AI in Cyber Policies | Cyber Threat Management AI | ML in Fraud Prevention
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What happened to Claude Fable 5 and Mythos 5?
The US Department of Commerce ordered Anthropic to suspend access on June 12, and the company took both models offline worldwide. Commerce lifted the controls on June 30, with restoration starting the next day.
Why were the models taken offline?
According to the source material, the trigger was a disputed security concern involving alleged jailbreak prompts that could produce cyberattack-useful output. Anthropic disputed the severity of that claim.
Who was affected by the shutdown?
The outage hit access through direct Claude APIs and major cloud providers, including AWS Bedrock, Google Cloud and Microsoft Foundry. The source material says enterprise users in finance, healthcare, SaaS and infrastructure were among those exposed.
What changed before access returned?
Anthropic agreed to new security-risk protocols, malicious-activity reporting and future-release procedures. A new safeguard reportedly blocked the alleged jailbreak about 93% of the time in CAISI testing.
What should AI builders watch now?
Builders should watch whether government review becomes a standing step for frontier releases and whether the August benchmark deadline creates uniform rules. The main operational lesson is to plan for portable model access.
Source: Thorsten Meyer AI
