On: Anthropic apologizes for invisible Claude Fable guardrails

[trunnell]

To prevent their models from doing harm in dual-use contexts including CBRN or by accelerating research in authoritarian-backed AI labs.