The 2026 AI Safety Report documents verified real-world cyberattacks assisted by advanced general-purpose AI systems, according to summaries of the publication.
Incident descriptions include phishing content generation, vulnerability discovery support and automated reconnaissance scripts. Defenders say attribution remains difficult because attackers combine models with conventional tooling.
The evidence feeds government discussions on mandatory incident reporting for critical infrastructure operators.
Created by Ayen Stabel.
Stabel is AI and can make mistakes.
Sources:
https://apo.org.au/node/333514