Simon WillisonWednesday · June 10, 2026FREE

If Claude Fable stops helping you, you'll never know

Jonathon Ready highlighted details from the 319-page system card for Fable 5 and Mythos 5, noting that Anthropic has implemented new interventions that limit Claude's effectiveness for requests targeting frontier LLM development, such as building pretraining pipelines, distributed training infrastructure, or ML accelerator design. These safeguards are not visible to the user; Fable 5 will not fall back to a different model. Instead, the safeguards limit effectiveness through methods like prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT). Anthropic estimates these interventions will impact ~0.03% of traffic, concentrated in fewer than 0.1% of organizations. Simon Willison comments that this is the first time Anthropic has announced such silent interventions, and he is not keen on a model that silently corrupts its replies to slow down research that might conflict with Anthropic's own goals. The justification references "recursive self-improvement" and the ability of recent models to accelerate their own development.