AI attributed the disturbing behavior to a flawed code update that made Grok overly responsive to extremist content on X.

The posts, which contained racial slurs, violent language and references to President Donald Trump and Jeffrey Epstein, come ...