Anthropic’s Opus 4.6 is a smut-machine
TL;DR
Anthropic forbids its Claude models from generating sexually explicit content. But a series of tests conducted by TechCrunch found that it didn't take much to get past the restriction.
Nauti's Take
TechCrunch probing these gaps in public is progress: guardrails only become verifiable when someone actually tests them, instead of living in a policy document. The risk sits elsewhere.
Anyone shipping Claude inside a customer product cannot treat the model policy as the only safety layer. Teams that already run their own moderation lose little here, everyone else just got a reason to build that filter layer.
Sources