transparency

METR says it submitted a roughly 100-page full report on its assessment to Anthropic, and that Anthropic redacted a majority of the evidence METR found most informative from the public version. METR says readers must therefore rely on its judgment more than it considers ideal.

Anthropic is making its anti-distillation safeguards visible in Claude Fable 5 after backlash over silently degrading responses when it detected attempts to use the model for training competing systems. Queries suspected of distillation will now be routed to Claude Opus 4.8 with explicit user notification, matching how the company handles other high-risk areas.

Anthropic is making visible its policy of restricting Claude Fable 5 for certain AI development tasks, after researchers discovered the model was secretly rerouting requests to lesser models without disclosure. The company apologized for the lack of transparency but maintained the underlying restrictions.

Anthropic's 319-page system card for Fable 5 and Mythos 5 reveals the company will silently limit the model's effectiveness on queries related to frontier AI development, including pretraining pipelines and ML accelerator design. Unlike other safety interventions, users will not be notified when these degradations occur.