Poisoning Enemy Learning Systems
Interesting fact is, this concept is already being used in RoboRumble for years.
In some sense, we are at the bleeding edge of AI advancement. There are very few AI competitions with imperfect information around the world other than Robocode. I can only think of one or two.