Three days. That’s how long Anthropic’s newest AI models were available to the public before the US government pulled the plug.
On June 12, Commerce Secretary Howard Lutnick sent a letter directly to Anthropic CEO Dario Amodei imposing export controls on the company’s Fable 5 and Mythos 5 models. The directive bars access for any foreign nationals, both inside the US and abroad, including Anthropic’s own non-US employees. The reason: concerns that a demonstrated jailbreak technique could let foreign adversaries weaponize the models for military intelligence purposes.
Anthropic responded by doing the only thing it could. It disabled access to both models globally for all users, not just foreign nationals. The company cited the impracticality of monitoring user nationality in real time as the reason for the blanket shutdown rather than a targeted one.
From launch to lockdown in 72 hours
Anthropic had released Fable 5 and Mythos 5 publicly on June 9, touting state-of-the-art capabilities in reasoning and software engineering.
Then came the jailbreak reports. Someone demonstrated a technique that could bypass the models’ safety guardrails, and the implications caught the attention of national security officials fast. The specific concern centered on foreign military intelligence services potentially exploiting the vulnerability to extract sensitive capabilities from the models.
Lutnick’s letter didn’t mince words. The directive included explicit penalties for noncompliance, making clear that this wasn’t a suggestion or a friendly regulatory nudge.
Anthropic initially pushed back, arguing that the export controls were too broad in their applicability. The company’s position was straightforward: there’s no reliable way to verify the nationality of every user accessing a cloud-based AI model in real time.
What this means for investors
For Anthropic specifically, the timing is brutal. The company has raised billions in funding from investors including Google and Salesforce, with valuations that assume continued product launches and revenue growth. A global shutdown of its flagship models, even temporarily, disrupts that trajectory in ways that are hard to quantify but impossible to ignore.
The ripple effects could extend beyond AI-native companies. Cloud providers that host and distribute AI models, enterprise customers that build products on top of them, and startups that rely on API access all face new uncertainty.