Your Discrete Graph Diffusion Model is Secretly Memorizing
Graph AI models learn to copy training data, hiding the problem until you look closely
Researchers discovered that graph diffusion models—AI systems designed to generate new networks—are secretly memorizing their training data. The memorization only becomes visible when you feed corrupted training graphs back into the model; standard tests miss it entirely. The model recovers training graphs nearly perfectly while performing much worse on new graphs it hasn't seen, and this problem worsens over training time regardless of when you stop.
Graph diffusion models are used to generate synthetic networks in drug discovery, social networks, and recommendation systems. If they're memorizing training data without obvious signs, they could leak sensitive information from real networks (like patient relationships in medical data) while appearing to generate novel graphs. This finding means current checks for whether these models actually learn general patterns are insufficient, and developers need better ways to spot hidden memorization before deploying them on confidential data.