Syncophancy is implicitly becoming a critical issue in using AI scientists or in general using AI in science. I think what this means is that scientists should remain critical with AI outputs, and also we need tools to reliably evaluate AI generated research works.
