Behind Fable 5's Comeback, Anthropic's Relationship With Washington Sours -- Could Europe Become a 'Safe Harbor'?

This week's AI news is dominated by Anthropic, starting with the best news of all: Fable 5, previously pulled offline, is back. It's currently the world's most powerful AI model, though how it compares to OpenAI's new GPT 5.6 Sol remains to be seen. Fable 5 is included in standard Claude plans until the 7th; after that, it reverts to usage-based billing and credits, like the API, and pricing rises significantly -- so the advice is to make the most of the remaining days.

The discussion opens with one figure's extreme proposal (dismissed as completely insane), before turning to the broader point: while it's hard to imagine Dario Amodei personally making such decisions, Anthropic's relationship with the US government overall keeps deteriorating. If Washington keeps pushing Anthropic, the company may be forced to look for an alternative base -- and Europe is signaling it stands ready as a 'safe harbor.'

But Europe also needs to be honest about its own serious AI governance problems, and needs its own DeepSeek- or Anthropic-caliber player. Mistral, often cited as Europe's champion, is assessed harshly: roughly 30% of it is American capital, and it hasn't proven as competitive as the other major players.

The segment closes with a thought experiment: imagine if someone had tried to ban the automobile when it first appeared, citing rights concerns. The same thing happened historically with the printing press -- cities that adopted it grew 60% faster than those that didn't, and eventually everyone adopted it anyway.

On a specific company (implicitly Palantir), opinions can reasonably differ, but simply exercising a veto in the absence of any alternative or competition is a mistake, the host concludes. The ideal would be a choice among three or four providers -- but that's not the reality today, and judgment should account for that.