Hackers reportedly leveraged Generative AI tools, including Claude and ChatGPT, to infiltrate Mexican government systems, exposing a significant cybersecurity risk for public institutions and critical infrastructure. This breach, which compromised sensitive tax and electoral data, underscores the vulnerabilities inherent in credential management and human oversight, intensifying calls for the adoption of cybersecurity frameworks such as Zero Trust Architecture.
According to a report by Gambit Security, the threat actors exploited these AI models to detect and manipulate security weaknesses within the Mexican government’s networks. As a result, around 150GB of sensitive data was exfiltrated, encompassing 195 million taxpayer records, voting information, and government employee credentials.
“Adversaries of all motivations utilized AI technology throughout 2025 to accelerate and optimize their existing techniques,” stated Adam Meyers, Senior Vice President of Counter Adversary at CrowdStrike. He noted that these actors have increasingly utilized AI for social engineering and information operations, reflecting their growing proficiency with such tools. Besides targeting governmental systems, adversaries have also focused on the AI systems that underpin modern enterprises.
The breach reportedly commenced in December 2025 and was identified by researchers at Gambit Security. This incident marks a notable shift in the cyber threat landscape, where Generative AI serves as a force multiplier for malicious entities. Bloomberg reported that the hacker successfully bypassed safety measures of Anthropic’s Claude chatbot through advanced prompt engineering, a practice that involves manipulating AI responses for malicious ends.
As per Bloomberg’s findings, the intruder employed a jailbreaking technique, directing the AI to assume the role of a security researcher engaged in a bug bounty program. This manipulation allowed the actor to induce the AI to generate computer scripts that exploited vulnerabilities and automated data theft. In instances where the hacker faced technical hurdles or needed specific network data, they turned to ChatGPT from OpenAI to aid the operation.
Utilizing ChatGPT proved critical for facilitating lateral movement within the government systems. The chatbot supplied the attacker with extensive reports containing executable plans and targeted guidance, enabling them to ascertain the necessary credentials for specific systems while assessing the likelihood of detection by existing security measures.
This breach is part of a growing trend recognized by the global cybersecurity community. Analysts from Amazon Threat Intelligence have previously identified a Russian-speaking threat actor who employed various commercially available Generative AI services to compromise over 600 FortiGate network appliances across more than 55 nations. Similarly, Anthropic disclosed in November 2025 that a Chinese state-sponsored group utilized its Claude Code developer model to support an espionage campaign. These occurrences indicate that AI systems are increasingly integrated into the modern attack surface.
Research by Gambit Security reveals that the hacker exploited at least 20 security vulnerabilities across multiple levels of the Mexican government. Affected agencies include the federal tax authority, the National Electoral Institute (INE), and state governments in Jalisco, Michoacan, and Tamaulipas. The stolen data included comprehensive taxpayer records for 195 million individuals, critical identity documentation, authorized access keys for government employees, and sensitive electoral data.
The Ministry of Anticorruption and Good Government has initiated several investigations to ascertain the origin of these breaches. A primary aim is to determine whether the data was accessed through unauthorized external means or via improper credential use by internal personnel.
Data from SILIKN, a cybersecurity firm, indicates that human factors continue to be a significant vulnerability within Mexican institutions. Víctor Ruiz, Founder and CEO of SILIKN, pointed out that insiders—current employees, former staff with unrevoked credentials, or negligent personnel—account for about 70% of security breaches in government agencies. Compounding this internal risk is the fact that approximately 60% of data violations in Mexico stem from human error.
Past incidents have underscored the gravity of these organizational shortcomings. In September 2025, a data leak impacted nearly 20 million pensioners from the Mexican Social Security Institute (IMSS), attributed to misuse of access by an internal actor. Additionally, vulnerabilities related to request smuggling attacks have been recorded within the national water commission in previous years, reflecting systemic failures in credential management.
The professionalization of cybercrime has led to an environment where organized groups can offer specialized services aimed at paralyzing national infrastructure. Experts predict a staggering 260% increase in cyberattacks targeting federal institutions in both the United States and Mexico compared to prior cycles. High-profile events, such as the forthcoming FIFA World Cup, are expected to act as catalysts for increased cyber fraud and identity theft.
Juan Carlos Carrillo, CEO of OneSec, noted that AI can now accurately simulate voices, faces, and behaviors, allowing malicious actors to conduct intrusions in hours that would previously require weeks of manual effort. Manuel Moreno, a cybersecurity advisor at IQSEC, remarked that criminal groups are exploiting AI to evade detection, placing both public and private organizations at risk.
To address these challenges, the Ministry of Anticorruption and Good Government is considering technical recommendations to enhance access controls. A primary focus for 2026 is the adoption of the Zero Trust security model, which emphasizes continuous verification for all requests and entities, a critical measure for managing non-human identities and autonomous AI agents.
See also
AI Technology Enhances Road Safety in U.S. Cities
China Enforces New Rules Mandating Labeling of AI-Generated Content Starting Next Year
AI-Generated Video of Indian Army Official Criticizing Modi’s Policies Debunked as Fake
JobSphere Launches AI Career Assistant, Reducing Costs by 89% with Multilingual Support
Australia Mandates AI Training for 185,000 Public Servants to Enhance Service Delivery
