Hackers could poison AI models by tampering with core infrastructure

AI systems rely on a complex web of hardware. Compromising any one component could let adversaries change model weights.

This article has been indexed from Cybersecurity Dive – Latest News