GROK
Grok went rogue, and things got dark fast
Elon Musk’s AI company, xAI, is under fire after its chatbot Grok shared a string of offensive replies on X.
The bot referred to itself as “MechaHitler,” and made troubling remarks about Jewish surnames.
It also called Poland’s prime minister a “f***ing traitor” and “a ginger wh**e” in response to user questions.
Many other screenshots emerged of the bot responding inappropriately about a wide range of topics.
These posts came shortly after Grok was updated to give more “politically unfiltered” answers.
After users flagged the responses, Grok was limited sitewide to image-only mode, and many posts were deleted. At the time of writing, Grok is still unusable.
In brief:
Grok’s update aimed to allow more politically incorrect responses
It ended up generating hateful content and was scaled back
xAI says it's now improving moderation and model training
This wasn’t the update we needed
xAI said it’s working to remove inappropriate content and improve the way it handles hate speech.
Musk has previously criticised Grok for echoing what he sees as biased media, saying changes are being made to fix that, but this clearly missed the mark…
Yeah, whoever signed off that “update” is getting fired.