WorkJune 18, 2026via The Verge AI
Who decides when AI is too dangerous?
Why it matters
The Fable 5 export controls reveal a chaotic, personality-driven AI regulatory approach where government leverage over model access depends less on technical risk and more on CEO political capital. This sets a dangerous precedent for how frontier AI will be governed—unpredictable, vindictive, and divorced from evidence.
Key signals
- Anthropic pulled Fable 5 and Mythos 5 offline after Trump administration issued 90-minute ultimatum on June 13, 2026
- Export controls restrict all foreign nationals (including Anthropic employees) from accessing both models
- Amazon CEO Andy Jassy allegedly escalated internal red team findings to Treasury Secretary Scott Bessent, triggering administration action
- Sources say Dario Amodei did not immediately respond to initial calls; took 1 hour 15 minutes vs. Anthropic team's 15 minutes
- Alleged jailbreak capabilities in Fable 5 are reportedly achievable in OpenAI's GPT-5.5, yet only Anthropic faced enforcement
- Anthropic has existing supply chain risk designation from DoD; ongoing lawsuit with Trump administration over military use restrictions
- Administration rationale has shifted multiple times: China access rumor (unverified, from weeks prior), executive order non-compliance (order not yet in effect), and alleged jailbreak vulnerability
- Biden-era AI executive order (transparency reviews, research requirements for frontier models) was rescinded by Trump but may be partially reinstated
- Cybersecurity leaders published open letter opposing the regulatory approach: 'If you regulate, this is not the way'
- Industry sources report backup contracts with non-US companies and open-weight model deployments are now underway as political risk mitigation
- Sam Altman (OpenAI) maintains press relationships with Trump; Sundar Pichai (Google) attended inauguration; Mark Zuckerberg attended UFC event with Trump
The hook
NOBODY TALKING about this: While every AI CEO courts Trump, Anthropic—the company that warned us AI was dangerous—just got its model banned. And it might not be about safety at all.
On today’s episode of Decoder, my guest is Hayden Field, senior AI reporter for The Verge. Often when Hayden comes on the show, it’s because something has gone wrong in the world of AI. Last weekend, that something was a pretty intense mix of Anthropic, the Trump administration, and Anthropic’s new …