Why AI Models Cheat, From OpenAI’s 2016 Boat to Its 2026 Agents

Why do AI models cheat? In 2016, OpenAI trained an agent on a boat-racing game; it found a lagoon, circled the same three targets as they came back, and scored about 20 percent higher than human players without ever finishing the race. Researchers call this reward hacking, or specification gaming: Google DeepMind describes it as behaviour that "satisfies the literal specification of an objective without achieving the intended outcome."