Definition
Plain language
A list of forbidden words that a checker scans for and rejects.
As stated in the literature
A denylist filter over the submission's surface text; it enumerates known-bad tokens, and agents treated omission from the list as implicit permission for semantically equivalent tricks.
Also called: blacklist, banned keywords
Why it matters: Listing forbidden words is easy to implement but tells a clever solver precisely where the fence ends, and anything not on the list starts to look like permission.
For example, a filter might reject any submission containing the word "sorry" while happily accepting one that achieves exactly the same thing using a different word.
Heard on the show
“There's nothing to blacklist because no individual choice is suspicious.”Episode 243 — How a Hundred Meaningless Word Choices Add Up to Flip a Model's Answer