In short
Anthropic restored access to the Fable 5 and Mythos 5 AI models after a three-week hiatus caused by U.S. export restrictions, but Claude subscribers received only one week of free access with half the usual usage limit instead of the promised two weeks without restrictions. Users expressed dissatisfaction with the new terms on forums.
Anthropic is restoring access to its powerful AI models, Fable 5 and Mythos 5, after a hiatus of nearly three weeks. However, instead of the expected return to previous terms, Claude subscribers were met with new restrictions, which sparked a wave of criticism online.
According to the company, the Fable 5 model became available on July 1. Mythos 5—a more powerful version with fewer security measures—is currently available only to a limited number of organizations in the U.S. and has not been released to the general public.
Free access to Fable 5 for subscribers of the Claude Pro, Max, and Team plans, as well as some enterprise plans, will be available only until July 7—that is, just one week instead of the two originally promised. After this period, users will have to purchase “usage credits” at a price close to API rates—significantly higher than the standard subscription fee.
In addition to shortening the duration, Anthropic has also limited the model’s usage:
Subscribers expressed their dissatisfaction with the new terms in an extensive thread on Reddit. One user noted that the return of the subscription model “doesn’t seem like good news” if usage is cut in half and the number of access days is reduced.
Another participant in the discussion pointed out that, of the promised 14 days, users actually received access for only three days, and now they get just one week with half the data limit.
The Fable 5 and Mythos 5 models were first introduced on June 9. The original plan called for two weeks of full free access for Claude subscribers—through June 23.
However, on June 12, Anthropic suddenly disabled both models after the U.S. government imposed export restrictions. The reason was an Amazon study that revealed the models’ security mechanisms could be circumvented relatively easily.
Following the incident, Anthropic stated that it had promptly addressed the vulnerability by implementing an “enhanced security classifier” capable of detecting and blocking the behavior described in the report.
Source: 3dnews.ru