Amazon CEO Andy Jassy raised security concerns about Anthropic’s Fable 5 AI model directly with US Treasury Secretary Scott Bessent and White House AI policy czar David Sacks — triggering a Commerce Department export control directive that forced Anthropic to shut down its two newest and most powerful models globally, just three days after they launched.
The models, Fable 5 and Mythos 5, had been available to Pro, Max, Team, and Enterprise subscribers since June 9, 2026. On June 12 at 5:21 PM ET, Anthropic received a letter from Commerce Secretary Howard Lutnick ordering access suspended. The directive was sweeping: it covered “any foreign national, whether inside or outside the United States, including foreign national Anthropic employees.” Rather than build selective per-user compliance, Anthropic disabled both models globally.
The API went dark at 6:59 PM PT that same evening, tracked in real time by developer Simon Willison.
What Triggered the Crackdown
According to multiple reports, Amazon researchers used prompts to get Fable 5 to provide information relevant to cyberattacks. Jassy shared these findings with senior US officials before the government moved. Anthropic pushed back publicly and directly. In an official statement, the company said:
“To date, the government has only given us verbal evidence of a potential narrow, non-universal jailbreak, which essentially consists of asking the model to read a specific codebase and fix any software flaws.”
The company also stated the same capability exists in GPT-5.5 and is routinely used by defensive security professionals:
“Our understanding is that one potential jailbreak was shared with the government. We have reviewed the report and validated that the level of capability displayed there is widely available from other models (including OpenAI’s GPT-5.5), and is used every day by the defenders who keep systems safe.”
Why Amazon’s Role Is Extraordinary
Amazon is not a neutral party. The company has committed to $100 billion in AWS cloud spending as part of its strategic investment in Anthropic. Jassy’s company is simultaneously Anthropic’s biggest infrastructure partner and the entity that reportedly triggered federal action against Anthropic’s own flagship products.
The Commerce Department invoked export control authorities — the same Export Administration Regulations (EAR) used to block Nvidia chip exports to China. Applying that framework to a cloud-hosted AI model is a significant, potentially precedent-setting move that every AI lab will be watching closely.
What It Means for Users
Fable 5 and Mythos 5 featured a 1 million token context window and always-on adaptive thinking — the most capable models Anthropic had ever shipped. They are now unavailable globally. All other Claude models — Opus 4.8, Sonnet 4.6, Haiku 4.5 — remain fully accessible.
Anthropic said on June 12 it would “share more details over the next 24 hours.” As of June 15, both models remain offline. Negotiations between Anthropic and the Commerce Department are underway (see update below). For affected users, the immediate alternatives include Claude Opus 4.8, GPT-5.5, Gemini 3.5 Pro, and — for coding tasks — the newly released GLM 5.2 from China’s Zhipu AI, which launched the same week under an MIT license with no export controls and a 1M token context window.
June 15 Update: Negotiations Underway
On June 15, the Commerce Department opened discussions with Anthropic about conditions under which the models might be restored. According to sources familiar with the talks, the government is considering two possible frameworks: a tiered access model that would allow verified US citizens and domestic enterprise customers to retain access while blocking foreign nationals, and a monitoring option that would require Anthropic to log and report usage patterns flagged by an automated compliance system.
Anthropic has not publicly confirmed the negotiation details, but the company’s public posture has shifted from pure pushback to cautious engagement. A full resolution — or formal rejection — is expected within days, making the Fable 5 and Mythos 5 shutdown one of the fastest-moving AI regulatory disputes in US history.
Our Take
Andy Jassy raising concerns about his own company’s biggest AI investment to federal regulators is the most unusual corporate power move of 2026. Whether Jassy acted out of genuine national-security concern or seized an opportunity to limit a competitor while appearing responsible, the result is the same: Anthropic’s most advanced models are offline, Amazon’s own AI offerings are unaffected, and the US government has demonstrated it can pull a commercial AI model from the market in under six hours. The commercial stakes could not be higher — Anthropic’s planned IPO is targeting a nearly trillion-dollar valuation, and this shutdown is a material risk to that timeline.
Frequently Asked Questions
Which Anthropic models are banned?
Fable 5 and Mythos 5 only. Claude Opus 4.8, Sonnet 4.6, and Haiku 4.5 are unaffected and fully available.
Why did Amazon report a flaw in Anthropic — its own investment?
Amazon has not publicly explained its reasoning. Sources say Jassy raised concerns with US officials about a jailbreak that could expose cyberattack-relevant information. Anthropic disputes the severity and notes the same capability exists in GPT-5.5 and is used daily by defensive security teams.
Will Fable 5 and Mythos 5 come back?
Anthropic is actively negotiating with the Commerce Department. As of June 15, two frameworks are under discussion: tiered access for verified US users and an AI-monitored compliance option. A decision is expected within days — bookmark this page for updates.
Is this the first time the US government has banned a commercial AI model?
Applying EAR export controls to suspend a cloud-hosted AI model for all foreign nationals globally — including a company’s own employees — is unprecedented.
