Hopefully AIs will do that because it is not difficult to imagine that bad actors will try to follow the same logic as commercial/political/etc. poisoners by in

Hopefully AIs will do that because it is not difficult to imagine that bad actors will try to follow the same logic as commercial/political/etc. poisoners by introducing vulnerabilities to code with a hope that an LLM will pick those up in the next round of learning.