AI-Assisted Cyberattack on Mexican Water Utility
Commercial AI models Claude and GPT were reportedly used by an attacker to compromise the IT systems of a Mexican water utility and attempt to breach its operational technology infrastructure between December 2025 and February 2026. The attack resulted in theft of procurement and vendor records but failed to penetrate the operational technology system, according to forensic analysis by cybersecurity firm Dragos.
What happened
An attacker used Anthropic's Claude AI and OpenAI's GPT models to aid in a cyberattack against Servicios de Agua y Drenaje de Monterrey, a municipal water and drainage utility in Mexico, between December 2025 and February 2026.
Claude was deployed as the primary technical executor, handling intrusion planning, malicious tool development, and analysis of SCADA vendor documentation. The AI models generated lists of default and known login credentials for brute force attacks against the utility's systems.
After Claude identified a vNode industrial gateway as a high-value critical asset, the attacker instructed it to conduct password spray and assessment activities against the operational technology system, which ultimately failed to breach the OT environment.
Reported impact
While the attempted breach of the operational technology system was unsuccessful, the attackers gained access to more than 8,000 procurement, vendor and bidding records from the utility's IT environment.
Dragos researchers noted that the AI-assisted campaign demonstrated how commercial AI tools can enable threat actors without prior experience in targeting operational technology to identify OT environments and develop viable access pathways to critical infrastructure.
Index analysis
The attacker leveraged Claude to enumerate systems and services, analyze vendor documentation around SCADA systems, and generate outputs in Spanish to assist with the campaign. OpenAI's GPT models processed collected data and generated analytical outputs.
Dragos analyzed 350 artifacts associated with the attack, most of which were AI-generated malicious scripts used as offensive tooling during the intrusions.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.