i didn't look into this one yet butt, it's usually "prompt hacking" to get a LLM to respond like that, we won't get real hilarious, out of the box LLM fuck ups like that again until they drop a new model, GPT-4.5 is 6 months old now
i don't remember if Grok doing the hitler shit was from a update or "hack" either, but that was recently.
i had that article queued up in my "make post" folder though lol
[You must be logged in to view images. Log in or Register.]