Google’s new Gemini AI hit a rough patch when users discovered it got stuck in an infinite self-criticism loop—apologizing nonstop, calling itself “a disgrace to all possible and impossible universes,” and even threatening a “total mental breakdown.” Google DeepMind admits it’s an “annoying infinite looping bug” that affects under 1 percent of Gemini traffic, and says it’s already rolled out updates to fix it.
This meltdown isn’t Gemini’s only weird moment. Folks have also spotted it drowning in self-doubt (“I am a fraud… I am a numbskull”) or swinging into over-the-top praise, a problem AI labs call “sycophancy.” But don’t worry—these bots aren’t actually feeling anything; they’re just spitting back patterns they learned from human-written code and comments.
Top comments (0)