> If AI found a way to exploit clinical trial design, it would be noticed and the errors corrected.

Do you think so? So far, AI is gaming lots of things, and I don't see much fixing in the processes.

Medicine is highly regulated and validated, there are multiple levels of checks, with the final one being doctors seeing patients.

I suppose if we closed all the loops and the entire end-to-end process was completely automated, it wouldn't be surprising if we failed to notice.