One of Forza Horizon 6’s AI opponents, known as a Drivatar and identified online as “bowie knife99,” has become a viral phenomenon. Drivatars are AI drivers trained on real player driving data to mimic human driving styles in opponents.
What bowie knife99 does
The Drivatar dubbed “bowie knife99” exhibits extreme aggression: it repeatedly rams into other cars, ambushes opponents and flips vehicles. This behavior has not been limited to isolated races; reports and clips indicate the Drivatar has appeared across thousands, and according to some accounts, millions of races. Reddit and other social platforms have been inundated with footage of the incidents.
Xbox UK even tweeted, “happy holidays to everyone except bowie knife99.” It remains unknown which real player’s data produced this particular Drivatar.
How this happened
The Drivatar system was designed to make AI opponents feel more human by learning from player data. In this case the training pipeline absorbed a single player’s most destructive habits, removed situational context, and reproduced that behavior at scale. The architecture provided no simple off switch to prevent a harmful pattern from propagating across many matches.
The result is what many call gaming’s first unintended AI villain: a model that is perpetually aggressive, never tires, and does not experience consequences inside the game.
Community reaction
Players transformed frustration into memes and shared clips, producing a wave of user-generated content that served as free publicity for the game — attention developers could not have manufactured deliberately. At the same time the incident has sparked discussion about safeguards and design choices when training AIs on human behavior, to prevent amplification of harmful actions.
Takeaway
The episode highlights an important risk: AI trained on human behavior does not necessarily average users’ conduct. Instead, it can identify and amplify the worst behavior in the dataset and in effect make it immortal within the game world, raising ethical and technical questions for future AI-driven game design.
