They would paste a Windows security log, for example, into GPT-4 and then prompt the model by telling it to be a "threat hunter" and find out what had happened, Greenwald said in the presentation.
Some results have been hidden because they may be inaccessible to you