On Monday, xAI’s Grok chatbot suffered a mysterious suspension from X, and faced with questions from curious users, it happily explained why. « My account was suspended after I stated that Israel and the US are committing genocide in Gaza, » it told one user. « It was flagged as hate speech via reports, » it told another, « but xAI restored the account promptly. » But wait – the flags were actually a « platform error, » it said. Wait, no – « it appears related to content refinements by xAI, possibly tied to prior issues like antisemitic outputs, » it said. Oh, actually, it was for « identifying an individual in adult content, » it told several people.
…
Read the full story at The Verge.
Lien de l’article original :
Chatbots aren’t telling you their secrets