Patrick Boyle
June 20, 2026
Anthropic Forced to Shut Down Claude Models in 90 Minutes
TL;DR
The US Commerce Department ordered Anthropic to shut down its Claude Fable and Mythos models globally within 90 minutes over alleged security concerns, raising questions about government overreach and the economic viability of US-dependent AI businesses.
Key Takeaways
- 1Anthropic received a Friday evening shutdown order claiming a jailbreak vulnerability, though the actual technique was simply asking Claude to read and fix code—its intended function.
- 2Amazon CEO Andy Jassy, whose company invested $13 billion in Anthropic and receives $100+ billion in AI computing contracts from them, reported the alleged vulnerability to the White House.
- 3Unable to verify which users were foreign nationals within 90 minutes, Anthropic disabled their flagship products globally, affecting major financial institutions that had committed to using Claude.
- 4The export ban accelerated Chinese AI adoption, with Deepseek and other open-source models now 60 times cheaper and immune to US government kill switches.
- 5The shutdown threatens the $965 billion valuation model that assumes global market dominance, as companies increasingly adopt foreign open-source alternatives.
Notable Quotes
“Until further notice, you must submit an application for an individually validated license prior to the export of the Mythos or Fable models to any destination worldwide or to any foreign person wherever they're located.”
— US Commerce Department letter
“if this standard were applied across the entire industry, it would essentially halt all new model developments for all frontier model providers”
— Anthropic statement
“if the United States from one day to the next can turn off the switch, it would damage the multi-trillion dollar companies leading the AI arms race”
— French President Emmanuel Macron
“It's not uncommon for governments to seek our council on potential security risks. When they occur, we don't share the details of these discussions.”
— Amazon statement
Chapters
1. The Friday Night Shutdown
Anthropic became the first AI company in history to have its flagship products forcibly disabled by the US government on a Friday evening with just 90 minutes notice, while the company was in confidential IPO filing with a near-trillion-dollar valuation.
2. The Alleged Security Threat
White House AI adviser David Saxs claimed a dangerous jailbreak was discovered in Claude Fable 5, but Anthropic revealed the technique was simply requesting the model to read code and fix bugs—its core advertised function.
3. The Cybersecurity Backlash
Over 100 cybersecurity executives from Google, Zoom, and other firms signed an open letter opposing the shutdown, noting that Claude's safety measures were so aggressive they refused innocuous tasks and that security professionals needed these tools.
4. Amazon's Complex Role
Amazon CEO Andy Jassy, whose company invested $13 billion in Anthropic and receives over $100 billion in AI computing contracts, was the unnamed 'trusted partner' who reported the vulnerability to the White House.
5. International Fallout
The shutdown scrambled global AI access for financial institutions, prompted European leaders to propose a $490 billion technological sovereignty initiative, and directly accelerated Chinese AI adoption.
6. The Rise of Chinese Alternatives
Chinese firms like Deepseek and Jiepu launched competitive open-source models that are 60 times cheaper, self-hostable (immune to US kill switches), and became four of the five most popular models on Open Router within weeks.
7. Valuation Unraveling
The shutdown undermines Anthropic's $965 billion valuation, which depends on global market dominance; if companies can access cheaper Chinese open-source alternatives immune to US regulation, profit margins collapse and the monopoly assumption fails.
8. Anthropic's Own Policy Irony
Days before the shutdown, Anthropic CEO Dario Amodei published an essay urging the government to build transparent regulatory processes for AI safety—exactly what he then experienced without transparency or technical justification.
Key People & Entities
- Dario Amodei
- CEO of Anthropic who received the Commerce Department shutdown letter
- Andy Jassy
- CEO of Amazon; reported the alleged Claude vulnerability to the White House
- David Saxs
- White House AI adviser who publicly justified the shutdown
- Howard Lutnik
- US Commerce Secretary who signed the shutdown order
- Sam Altman
- CEO of OpenAI; attended G7 summit lunch discussing AI opportunities
- Demis Hassabis
- CEO of Google DeepMind; attended G7 summit lunch discussing AI opportunities
- Emmanuel Macron
- French President who criticized the shutdown as damaging to AI competition
- Anthropic
- AI company forced to shut down Claude Fable and Mythos models globally
Glossary
- Deemed export
- Legal mechanism treating access granted to a foreign national inside US territory the same as physically exporting technology to their home country.
- Informed letter
- Private notice from US government indicating export controls now apply to technology that previously did not require licensing.
- Jailbreak
- Technique to bypass an AI model's safety guardrails and make it produce harmful outputs.
- Frontier AI models
- Advanced large language models representing the cutting edge of AI capability, typically costing billions to train.
- TAM (Total Addressable Market)
- The total market opportunity available to a company; used to justify valuations in venture capital.
- Constitutional AI
- Anthropic's AI safety framework designed to make models refuse harmful requests by adhering to a set of constitutional principles.
- Defense in depth
- Security strategy using multiple layers of protection so that if one layer is compromised, others remain intact.