The Once And Future Fable #2

TL;DR

• 1

  A narrow jailbreak triggered a global takedown: The speaker says Anthropic warned in advance that limited jailbreaks were possible, and that the reported outputs were no worse than what GPT-5.5 can already produce normally.

• 2

  Amazon's call appears to have set the crackdown in motion: According to Axios reporting cited here, Amazon contacted the administration Thursday night about a Mythos jailbreak, and calls from Amazon plus at least five other companies helped lead to Fable being shut down by Friday night.

• 3

  Anthropic says it got a 90-minute ultimatum with almost no details: The recap leans on FT, Axios, and Politico to argue that Dario Amodei asked for specifics, tried to explain the issue was narrow, and was punished for not instantly agreeing to pull the model.

• 4

  The speaker thinks this was more about vibes than technical evidence: He repeatedly argues the White House reacted to perceived disrespect, politics, and communication style, citing the 'wellness retreat' claim, complaints about Anthropic's 'lack of seriousness,' and hostility toward people seen as aligned with Chris Krebs or Democrats.

• 5

  Cybersecurity leaders say the move hurts defenders more than attackers: A letter cited via Axios, with names from Adobe, Zoom, and Sophos, argues Fable's reported capability was basic defensive bug-finding that can also be replicated on GPT-5.5, Opus, Sonnet, and Chinese models like Kimi 2.7.

• 6

  The real alarm is the precedent: AI is licensed now, just informally: Borrowing Dean W. Ball's framing, the speaker says the US now has an ad hoc licensing regime for frontier AI, one enforced through executive pressure and export controls without transparent standards or clear legal process.