Just saw that OpenAI shelved their adult chat feature for ChatGPT after some pretty sketchy test results came out. Apparently their age-checking system was misidentifying teenagers as adults like 12% of the time, which is... not great when you're trying to keep minors away from adult content. Their ethics council flagged it back in January saying it could be risky for vulnerable users, and honestly that makes sense. The whole thing got me thinking about how hard it actually is to build safe AI features, especially ones dealing with age verification and adult chat content. They're basically saying "we need to get this right before we launch" which I guess is the responsible move, but it also shows how many edge cases exist in AI safety. Anyone else following this stuff? Seems like every major AI move gets scrutinized now.