Double trouble: Hackers used both Claude Code and ChatGPT in a cybersecurity hack that lasted two and a half months.

What Happened

Nine Mexican government agencies fell victim to a sophisticated AI-powered cyber campaign that ran from December 2025 through mid-February 2026. Researchers at Gambit Security confirmed that attackers leveraged both Anthropic's Claude Code and OpenAI's GPT-4.1 to orchestrate the breach, maintaining persistent access for over two and a half months. This marks one of the first documented cases where multiple large language models were deployed in tandem to execute a prolonged attack against healthcare-adjacent government entities.

The attackers used AI tools to automate reconnaissance, craft social engineering attacks, and potentially generate malicious code — significantly reducing the skill barrier for sophisticated cyber operations. Gambit Security characterized the incident as a "wake-up call" for organizations that have not yet adapted their security posture to defend against AI-augmented threats.

Data Exposed

While specific data types compromised in this breach were not detailed in available reports, Mexican government agencies typically maintain:

• Personal health information (PHI) for public health programs
• Citizen identity records including national IDs and demographic data
• Employment and benefits information for government workers
• Administrative credentials and inter-agency communication records
• Healthcare eligibility and enrollment data for social services

The duration of the breach — over 75 days — suggests attackers had sustained access to exfiltrate substantial datasets.

Response & Remediation

Mexican authorities have not publicly disclosed the full scope of their response. Standard post-breach protocols would include:

• Immediate credential rotation across all compromised systems
• Forensic analysis to map the attack timeline and data accessed
• Network segmentation to contain lateral movement
• Enhanced monitoring for indicators of AI-generated attack patterns
• Vendor security reviews of third-party access points

The extended breach window indicates delayed detection — a common failure when organizations lack real-time security monitoring and anomaly detection capabilities.

Why It Matters

This incident represents a paradigm shift in threat actor capabilities. AI tools like Claude and GPT-4 enable attackers to:

• Automate social engineering with highly personalized phishing campaigns
• Generate polymorphic malware that evades signature-based detection
• Accelerate reconnaissance by processing vast datasets in seconds
• Lower the skill floor for executing advanced persistent threats

For independent healthcare practices, the lesson is clear: traditional compliance checklists are insufficient. Attackers no longer need specialized expertise to breach networks — they need an API key and persistence. Practices handling ePHI must assume adversaries now operate with AI assistance and adjust defenses accordingly.