Anthropic's new Claude Fable 5, the first public version of its restricted Mythos-class model, has drawn loud backlash from developers and researchers days after launch over three things. The model burns through paid usage limits roughly twice as fast as Opus 4.8, with one tester reportedly draining a $100 Max plan in under nine minutes, and Anthropic's own system card confirms it silently throttles its own performance, without telling the user, when it detects work on frontier AI development. The third complaint is a mandatory 30-day data-retention policy that critics say locks out GDPR-bound European companies and anyone needing zero retention, feeding a broader argument that Anthropic's safety framing masks moves that concentrate control and hit independent researchers hardest.
