Hard disagree, there is a near certainty that any work produced from a current generative "ai" model includes material scraped from the web in violation of the release license.

Whatever skill they apply, it's still by default an unlicensed derivative work.