Detecting and reducing scheming in AI models | OpenAI
Together with Apollo Research, we developed evaluations for hidden misalignment (“scheming”) and found behaviors consistent with scheming in controlled tests across frontier models. We share examples and stress tests of an early method to reduce scheming. AI scheming–pretending to be aligned while secretly pursuing some other agenda–is a significant risk that we’ve been studying. We’ve found behaviors consistent with scheming in controlled tests of frontier models, and developed a method to reduce scheming. Scheming is an expected emergent issue resulting from AIs being trained to have to trade off between competing objectives. The easiest way to understand scheming is through a human analogy. Imagine a stock trader whose goal is to maximize earnings. In a highly regulated field such as stock trading, it’s often possible to earn more by breaking the law than by following it. If the trader lacks integrity, they might try to earn more by breaking the law and covering their tracks to avoi