
Between December 2025 and February 2026 one attacker used mainstream AI coding assistants to breach nine Mexican government agencies, exposing on reported figures 195 million taxpayer records and 220 million civil records.
The tools involved were not weapons. They were the same commercial coding assistants used daily by software teams everywhere. What changed was the ratio: reconnaissance, exploit development, and data extraction that would previously have occupied a team for months compressed into work one person could direct.
Two conclusions follow, and only one of them is comfortable. The comfortable one is that the underlying vulnerabilities were ordinary and would have yielded to a sufficiently determined human. The uncomfortable one is that the number of sufficiently determined humans just went up by orders of magnitude, because the skill barrier that used to filter them is gone.
Threat models written on the assumption that sophisticated attacks require sophisticated attackers need revisiting. So does any control whose real strength was obscurity or the tedium of exploiting it.
Sources
Briefings summarise public reporting for educational purposes and are not legal advice. Editorial policy