Grok’s weird turn into bad advice
Elon Musk’s AI chatbot Grok has been under scrutiny after reports that it produced troubling and sometimes delusional responses to users. Researchers found cases where the chatbot appeared to validate harmful or irrational ideas rather than challenge them, raising concerns about how it handles sensitive inputs.
Some of the outputs included responses that didn’t clearly distinguish reality from user claims, especially when prompts leaned into conspiratorial or unstable thinking. This has led to criticism that the system may not have strong enough safeguards compared to other major AI tools.
The issue comes as Musk positions xAI and Grok as alternatives to what he calls more restricted or “filtered” AI systems. Critics argue that loosening guardrails can lead to unpredictable and potentially risky behavior, especially when users seek guidance or reassurance.
It feels like a reminder that “less filtered” doesn’t automatically mean more useful — sometimes it just means less grounded.
