They literally used the AI to do target selection in Iran and apparently moved half the vetting team to other projects, according to some articles that came out after the first wave of attacks. So while the AI wasn't making the final decision, they used it to accelerate the target selection process, but that is it literally making the first initial decisions.
They haven't actually admitted to it yet, but it is already being alleged, and I for one absolutely believe them because that is exactly the kind of dumb fucking shit they would do.
guys... grok is too powerful... grok is too world-changing...