Fable #6: The Return of the King
ai
Anthropic has restored Claude Fable after a brief shutdown. In June, researchers at Amazon discovered capabilities in Fable that triggered a government response—specifically, its ability to assist with debugging code in ways that concerned policymakers. The US government subsequently applied export controls to both Fable and its sibling model, Mythos.
Anthropic responded by refining its safeguard classifiers, dramatically expanding restrictions to refuse problematic requests in more than ninety-nine percent of cases. With those new guardrails in place, the government lifted export controls on Fable, and the model was restored worldwide on July first. However, the immediate safeguards are notably broad: routine coding and debugging tasks now fall back to Opus four point eight, Anthropic's smaller model. The company plans to refine these classifiers over coming weeks to reduce false positives and better distinguish legitimate requests from potential misuse.
Source: https://www.lesswrong.com/posts/r9HsHHSsfABhhxnYr/fable-6...
Listen to this story
Hear this and more stories in a personalized audio briefing.
Open The Chonkerton