The worst US decision in recent months. An opportunity for Europe?
Anthropic's Claude Mythos and Fable models have just been banned for all non-Americans. In practice, since Anthropic has no way to make the distinction (even within its own teams), that means the model is no longer available to anyone.
This ban is one of the worst US decisions in recent months.
In a few months (or a few weeks!), a model with equivalent or superior capabilities will be running in China or in my office, and no letter from the Department of Commerce will be able to unplug it.
This is probably an opportunity for Europe to position itself as the home of all the major open-source, deregulated models.
Unfortunately, I have little hope of seeing that happen, even though I would like it to.
Why will the models run elsewhere, and why am I even talking about my living room?
Quite simply because controls on models are much harder to enforce than controls on hardware chips.
Equivalent or superior capabilities are already emerging rapidly through open source, other labs or replication.
Once a capability exists, it spreads (very) quickly.
All my links, projects and social media: https://all.bo