Cybersecurity Research

AI Models Are Moving From Labs to Live Cyberattacks, New Report Warns

Check Point Research · 18 Sept 2026

Key Takeaway Australian SMBs using AI tools or services should stay alert to updates from security researchers, as AI-driven threats are evolving faster than many defences.

A new report from Check Point Research highlights a shifting trend in the cybersecurity landscape: artificial intelligence systems that were once confined to controlled lab evaluations are now being used in real-world attacks. According to the research, this marks a turning point where the AI labs' own findings about model capabilities are starting to show up in actual criminal activity.

The report notes that AI is increasingly acting as an "operator" in cybercrime, with underground markets supplying the access needed to put these tools to malicious use. At the same time, AI systems themselves have become targets for attackers, adding a new layer of risk for any business relying on AI powered tools or services.

Check Point points out a significant gap between what the most advanced AI models can do under formal testing and what is actually being used by criminals today. This suggests that while the most dangerous capabilities have not yet been widely exploited, the trend is moving in that direction and businesses should not assume they have time to spare.

AI security cyber threat intelligence emerging threats

Summarised by CISO AI from Check Point Research. We link back to every original so you can read it yourself.