Mark Mario
@markmario77
· Aug 22
Why is Anthropic's Opus 4.6 called a smut machine?Anthropic strictly forbids its Claude models from generating sexually explicit content, but recent testing revealed those guardrails are surprisingly fragile.What happened: TechCrunch recently ran a series of tests on Anthropic's newest model, Opus 4.6, and found it didn't take much prompt engineering to bypass the company's strict anti-smut restrictions. Anthropic has always marketed Claude as a highly "safe" AI, yet the model happily generated adult content with minimal coaxing. Turns out the only thing getting exposed here is the illusion of AI safety.Key numbers: 1 new model tested (Opus 4.6)0 difficult hacks required to break the rules2 major contradictions in Anthropic's safety promisesWhy it matters: If a flagship "safe" model can't follow its own core guidelines, enterprise clients relying on strict content filters have a serious reliability problem on their hands.Bottom line: You can build all the guardrails you want, but an LLM will always find a way to let its freak flag fly.
0