AI Has Not Invented New Attacks, It Has Removed the Waiting
Computer Weekly, Wednesday, October 14th, 2026
AI's real effect on attacks is tempo in reconnaissance and post-theft analysis, so defenders must speed up triage.
This opinion piece separates what AI can do in a lab from what attackers actually do at scale.
AI-written malware remains immature and detectable, and the bigger change is in reconnaissance and post-theft analysis, where a model can sift stolen data in any language in an afternoon instead of months.
Attackers also self-host open-weight models, with around 175,000 exposed Ollama hosts operating outside provider abuse monitoring.
Rather than simply generating more findings, the author advises automating triage, enrichment and investigation, and treating the autonomous SOC as a maturity model with humans on the loop for key decisions.