If a scientist uses AI to make strong claims in public about the scientific literature, and some references do not exist, and the generated text does not correcfly reflect the findings, should we inform them privately to fix the text or should we share this information publicly?