What the EU AI Act Actually Requires for Cybersecurity (And Where Enterprises Are Exposed)

**Executive summary:** This article outlines the EU AI Act's cybersecurity obligations for high‑risk and general‑purpose AI systems—naming threats such as data poisoning, model poisoning, adversarial examples, and confidentiality attacks—explains provider vs. deployer responsibilities, reporting timelines for serious incidents, GPAI documentation duties, and practical steps enterprises should take (inventory, classification, vendor demands, incident response, and AI literacy) to meet compliance and reduce exposure.