Dragos Documents First LLM-Assisted Strike on Water Infrastructure in Mexico (smallwarsjournal.com)

Cybersecurity firm Dragos has reported the first known cyberattack leveraging commercial AI models against a municipal water utility in Monterrey, Mexico. Between December 2025 and February 2026, attackers utilized Anthropic’s Claude for crafting intrusion plans and malicious scripts, alongside OpenAI’s GPT for data analysis and Spanish-language output generation. Although the operational technology infrastructure remained unbreached, the incident underscores how AI has revolutionized the threat landscape, making it easier for inexperienced actors to plan sophisticated attacks and refine their methods in real time. This development raises significant concerns within the AI/ML community, illustrating the dual-use nature of advanced AI technologies. The ease with which these commercial AI models can be harnessed for malicious purposes signifies a lowering of barriers to cyberattacks on critical infrastructure. Dragos warns organizations to adopt stringent remote access policies and robust authentication measures to safeguard against unauthorized access. Furthermore, the incident is reflective of broader governance challenges in Mexico, where systemic issues complicate cybersecurity preparedness, hinting that similar attacks might become more frequent as AI continues to evolve and integrate into various operational frameworks.