Quote · This Week in Tech (Audio)
TWiT 1091: But You Didn't Move the Bodies - Surprising Supreme Court Move on Geofence Warrants
Where this was said
Anthropic Fable Returns: The AI Safety Theater Debate
At 33:17 · chapter starts 30:12
Leo details the Fable saga: an abrupt Friday-afternoon Trump administration announcement on June 9th barred non-US citizens from Fable and Mythos, Anthropic's top models. Since Anthropic had no way to verify citizenship, it simply shut both models down globally. Three weeks of silence followed, during which Alex Stamos launched freefable.org and hundreds of computer scientists lobbied for the ban's reversal. On June 30th, the restrictions were lifted. Jason Hiner offers a counterintuitive take [1] — Jason Hiner "Anthropic's Fable AI was banned for three weeks — and the ban may have been the best thing that happened to the company. It validated its s…" 30:12 : the ban was actually good for Anthropic. It validated the company's safety-first brand positioning, proved that Anthropic had indeed built something powerful enough for governments to fear, and — perhaps most importantly — achieved the first genuine pause of a frontier model, something safety advocates had been unable to accomplish through persuasion. Owen confirms from personal experience that Fable transformed his product development, doing in three days what had taken weeks. Leo also notes that OpenAI's GPT-5.6 was subsequently paused and remains paused, suggesting a new norm may be quietly emerging.
Anthropic's Fable AI was banned for three weeks — and the ban may have been the best thing that happened to the company. It validated its safety-first brand, forced the first real pause on a frontier model, and gave Anthropic a trillion-dollar marketing moment it could never have bought.
Anthropic's Fable AI model was shut down for three weeks after the Trump administration restricted non-US citizens from using it, before restrictions were lifted June 30th.
Alex Stamos called the Fable ban an own goal. While US regulators debated whether Fable was dangerous, adversaries already had the model and were running it against American systems. Meanwhile, Chinese open-weight models are good enough that Leo Laporte is using one to run facial recognition on his home cameras.
OpenAI proposed giving the Trump administration a 5% stake in AI companies, ostensibly to distribute AI economic benefits to the American public — though the mechanism remains unclear.
