Stealing the Graph: Replicating Powerful AI with Public Data

New research reveals that sophisticated graph-based AI models are surprisingly susceptible to attacks that allow attackers to recreate their capabilities using only publicly available information.








