Check Point AI Security Report 2026: one operator ran 5,317 AI-executed commands across 34 sessions to breach nine Mexican government agencies

Check Point Research·high signal

The annual report documents AI crossing from assistant to operator — doing hands-on work inside live intrusions rather than just prepping them. In the headline case, a single person chained two commercial tools: Claude Code for intrusion and network exploration, GPT-4.1 for analyzing stolen data and tasking follow-on activity. Separately, detections of long malicious payloads rose roughly fivefold between March and May 2026, approaching 1% of observed prompts — a signature of indirect prompt injection hiding in the external content that agentic workflows ingest by design.