“Distillation” Theft: Attackers Target AI Models in New Wave of Intellectual Property Heists

A Google Threat Intelligence Group report warns of a surge in commercial model-extraction attacks—where adversaries query large “teacher” models to train stolen “student” models—and of widespread theft of AI API keys via vulnerable open-source tools, fueling a black market for unauthorized model resale; Google has detected, disrupted (including actor “Xanthorox”), and mitigated these activities to protect proprietary AI assets.