← Back to feed

Hackers could poison AI models by tampering with core infrastructure

AI · Security · 2026-11-01 22:17

Security firm Eclypsium warns that AI systems rely on complex layers of hardware and firmware, so attackers who compromise any single link in that chain could tamper with model weights and alter the outputs those models produce.

Source: www.cybersecuritydive.com


Auto-generated summary for reference only. RSS