Counter-Strike remains one of Valve's most popular games, with over one million multiplayer matches being enjoyed every single day. However, not all players are playing fair, and there's a range of cheats available to use. Therefore Counter-Strike requires cheat detection, the latest of which is an artificial intelligence.
Valve confirmed this in a response on Reddit to a question about "spinbots" and why they aren't auto-detected and kicked off the server. Spinbots, or spinhacks as they are also called, allow a player to automatically spin around quickly during play effectively giving them 360 degree vision. Combined with an aimbot, it becomes a highly effective way of killing anyone who happens to enter the players much-enhanced field of view.
According to Valve, automatically detecting such activity would just lead to "an arms race with cheat developers" who would find the limits of the detection and work around them. You also hit the issue of "is that a cheat or a very good player?" The other big problem is that running auto detection takes a lot of processing power, and ultimately is unworkable for a million+ matches every day, so an alternative method is required.
Counter-Strike already uses a system known as Overwatch to allow the Counter-Strike community to regulate itself using in-game replay reviews. So Valve took the next step and implemented an AI that takes advantage of the Overwatch system.
Using machine learning, the automated system trains continuously and over time can figure out the difference between a cheat and a highly-skilled player. An early version of this AI is already up and running with the results being "promising." Valve is now "going to continue this work and expand the system over time."
If the machine learning system continues to improve itself, Counter-Strike players should eventually have an AI working with them to weed out the cheats and remove the frustration of spinbots and aimbots from play automatically.