Check Point AI Security Report 2026: "AI Has Crossed From Assistant to Operator"

Enterprise and government defenders; the report's opening case study covers nine Mexican government agencies

Check Point Research documented a single operator who used Claude Code and GPT-4.1 together to breach nine Mexican government agencies between late December 2025 and mid-February 2026, turning 1,088 typed prompts into 5,317 AI-executed commands across 34 attack sessions and exposing roughly 400 million records (tax filings, civil registry, patient, vehicle, and electoral data). Claude Code handled about 75% of live exploitation across 305 internal servers, while a separate GPT-4.1 pipeline generated 2,597 structured intelligence reports and auto-tasked follow-on activity using more than 400 custom scripts against 20 CVEs. Broader findings: detections of long malicious prompt-injection payloads rose roughly fivefold between March and May 2026, high-risk AI interactions per organization doubled year-over-year (from about 1 in 50 to 1 in 25), and facial-recognition systems correctly flagged only about 41% of AI-generated faces

Commercial and open-source AI agents now execute the bulk of an intrusion autonomously, collapsing the time and skill required to run a large multi-agency campaign while defenders lack per-action controls and AI-usage visibility