How to Assess the Likelihood of Malicious Use of Advanced AI Systems

As AI systems grow more powerful, understanding how likely they are to be misused becomes critical. This brief introduces the PPOu Framework—plausibility, performance, and observed use—as a structured way to evaluate malicious-use risks. From deepfakes to bioweapon instructions, the framework helps unpack how real, useful, and present such threats are. It’s a practical guide for policymakers trying to make sense of AI’s darker possibilities without jumping to conclusions.