This is the first time I’m doing this, but it felt appropriate given the turn of events.
Not to be outdone by Anthropic’s blocking of cybersecurity, biology, and AI model researchers, the US government issued a directive to Anthropic Friday evening, saying Fable 5 and Mythos 5 posed a national security risk, and therefore no foreign nationals inside or outside of the US, including Anthropic employees, should be allowed access to either model. According to Anthropic, the rationale was that there was a possibility of jailbreaking the model (may have been referring to a publicly disclosed jailbreak that Anthropic disputed counted as defeating it), though Anthropic also claims the same vulnerabilities the US government may be referring to and using as the basis of the directive are found in other models, as well.
As the BBC also states, Anthropic and the US government are still in legal battles over whether or not US government employees are allowed to us Anthropic tools in the first place.
So, after all the drama of Fable 5’s release, debates over the effectiveness, price, when to use or not use it, and people maxing token usage after a single prompt (Yahoo! seems to have hit 1.8M tokens in one prompt on the model), nobody can access it at all.
Anthropic ends their memo with an apology to users, saying this is likely a misunderstanding and hoping to restore access soon.
With all of this, who’s ready to read some Fablehaven or play a game of Fable?