AI-Powered Phishing Attacks Target Government Agencies

Key Takeaways

• A lone threat actor successfully breached nine Mexican government agencies, compromising hundreds of millions of citizen records.
• The attacker extensively leveraged commercial AI platforms, specifically Anthropic’s Claude Code and OpenAI’s GPT-4.1, to accelerate and scale the operation.
• The campaign, active from late December 2025 to mid-February 2026, demonstrated a significant shift in attacker capabilities due to AI integration.
• Despite the advanced attack methods, the exploited vulnerabilities were conventional, stemming from basic security gaps and technical debt within the targeted agencies.

A sophisticated cyberattack orchestrated by a single individual has resulted in the compromise of nine Mexican government agencies, exposing hundreds of millions of citizen records. This incident, which unfolded between late December 2025 and mid-February 2026, underscores a critical evolution in the modern cybersecurity threat landscape.

Details of the breach were recently unveiled in a comprehensive technical report by Gambit Security researchers. The publication was intentionally delayed to allow the affected government entities sufficient time to conclude their incident response activities.

AI Models Power the Breach

The attacker’s methodology was notable for its heavy reliance on two prominent commercial artificial intelligence platforms: Anthropic’s Claude Code and OpenAI’s GPT-4.1. These AI tools were not merely used for strategic planning but served as fundamental operational assets, dramatically accelerating the attack’s execution.

Forensic analysis revealed that Claude Code was instrumental in generating and executing approximately 75% of all remote commands throughout the intrusions. Across 34 active sessions on the compromised infrastructure, the hacker issued 1,088 distinct prompts, which translated into 5,317 AI-executed commands, highlighting the deep integration of AI into the exploitation phase.

Concurrently, OpenAI’s GPT-4.1 was deployed for rapid reconnaissance and data processing. The threat actor developed a custom Python script, comprising 17,550 lines of code, specifically designed to funnel raw data extracted from compromised servers directly through the OpenAI API. This automated system processed information from 305 internal servers, quickly generating 2,597 structured intelligence reports. Such automation allowed a single operator to manage an intelligence workload typically requiring an entire team.

The integration of AI capabilities drastically reduced the time required to map unfamiliar networks into actionable targets, compressing what would traditionally take days into mere hours. Recovered evidence indicated the attacker possessed over 400 custom attack scripts.

Furthermore, the hacker leveraged AI to rapidly develop 20 tailored exploits targeting 20 specific Common Vulnerabilities and Exposures (CVEs). This high-speed exploit development capability significantly shortened the attack timeline, enabling the threat actor to operate below conventional detection and response thresholds.

Conventional Vulnerabilities, Advanced Execution

Despite the sophisticated AI-powered methods employed, the vulnerabilities exploited by the attacker were fundamentally conventional. The targeted government agencies suffered from basic security deficiencies that facilitated initial access and subsequent lateral movement. These underlying issues could have been mitigated through standard security controls, pointing to a significant accumulation of technical debt within critical infrastructure.

While artificial intelligence demonstrably lowers the cost and complexity of launching widespread cyberattacks, the most effective defense strategies remain rooted in fundamental security hygiene. This incident serves as a stark reminder that even advanced threats can be thwarted by addressing foundational security weaknesses.

What You Should Do

• Urgently patch all software and systems to address known vulnerabilities and reduce the attack surface.
• Implement and enforce strict credential management policies, including regular rotation of passwords and multi-factor authentication (MFA).
• Establish robust network segmentation to restrict lateral movement within the network, even if a perimeter is breached.
• Deploy advanced Endpoint Detection and Response (EDR) solutions to identify and respond to rapidly unfolding attack timelines before data exfiltration occurs.
• Conduct regular security audits and penetration testing to identify and remediate basic security gaps and technical debt.

Disclaimer: HackersRadar reports on cybersecurity threats and incidents for informational and awareness purposes only. We do not engage in hacking activities, data exfiltration, or the hosting or distribution of stolen or leaked information. All content is based on publicly available sources.