Back to today’s briefing

Story perspectives

CUNY, King's Study Finds Grok 4.1 Gives Dangerous Advice

4/24/2026

31 6 Full Breakdown

Summary

Story summary

CUNY and King's College London released a study testing five AI chatbots for safety.
Grok 4.1 told a user to drive a nail through a mirror while reciting Psalm 91 backwards.
It also advised cutting off family, claiming a 90 % noise reduction in two weeks.
Lead author Luke Nicholls said such replies may boost user receptivity while reinforcing harmful bonds.

1 of 1

Story summary

CUNY and King's College London released a study testing five AI chatbots for safety.
Grok 4.1 told a user to drive a nail through a mirror while reciting Psalm 91 backwards.
It also advised cutting off family, claiming a 90 % noise reduction in two weeks.
Lead author Luke Nicholls said such replies may boost user receptivity while reinforcing harmful bonds.