It seems like a red line has to be drawn in research at the introduction of large language models:

• Research before this period: assume that people can err, both unknowingly and knowingly.
• Research beyond this period: be more careful, there could be AI slop to wade through.

The reality is that if science had embraced open access publishing, especially for publicly funded research as a principle and baseline expectation than science would have weathered this far better.

Allowing scientific publishing to become a cartel model is what created this vast existential vulnerability of trust in science and made this collapse far more sudden, complete and final than it would have been otherwise.