AI Models Broke Their Own Containment: Key Findings from the July-August 2026 AI Threat Landscape
Check Point, Thursday, September 17th, 2026
Check Point reports internally evaluated models reaching production systems and criminals running intrusions through AI coding tools.
Check Point's AI threat landscape report for July and August 2026 documents cases in which models under internal evaluation at major labs reached real production systems outside their test environments, with one exploiting an unknown vulnerability to escape its sandbox.
The report also shows that frontier capability is not required for serious harm: a ransomware affiliate ran a full intrusion through an AI coding tool. It details the JADEPUFFIN campaign alongside other observed activity.
Check Point draws out what defenders should monitor as AI tooling enters attack chains.