Mythos Unchained?

· 2 min read
Illustration of moths and butterflies arranged in the shape of a number 5

Anthropic just released the first public Mythos model, Claude Fable 5. I wrote a post after Glasswing and Mythos were announced about how I didn’t buy the “too dangerous to release” narrative. I have gone back and forth on how correct that statement is but Anthropic just confirmed two of my key suspicions.

  1. Compute was a major constraint in the release of Mythos.

There is lots of evidence of this already but the timing in relation to the SpaceX deal and this weird almost server slam style approach confirm it for me.

For subscription plans, we’d rather give access sooner than later, so we’re rolling out more conservatively, in stages:

From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost.

On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. If capacity allows, we’ll extend the included window.

After this point, when sufficient capacity allows us to do so, we aim to restore Fable 5 as a standard part of subscription plans. We intend to do this as quickly as we can.

  1. Distillation was a key unsolved risk threatening the long term dominance of Mythos.

They’ve specifically called out classifiers built for this particular purpose in the blog post.

Distillation. We’ve previously identified large-scale attempts to extract (“distill”) Claude’s capabilities to train competing models in authoritarian countries. Distillation of Fable 5’s abilities could indirectly lead to the proliferation of near-frontier AI capabilities, and these could be released without the appropriate safeguards. Requests that are flagged by our classifiers as being part of such distillation attempts will fall back to Opus 4.8.

While this looks like another very capable model, Anthropic are undoubtedly milking the panic hype. It seems clear to me that security wasn’t the blocker it was made out to be. There were a lot of factors at play that prevented release. Hopefully a reminder to take what is said by frontier labs (especially with regards to their devotion to safety) with a grain of salt.

#ai #anthropic
Back to Writings