Mexico's Tax Authority

The organisation is known as Mexico's Tax Authority, with its headquarters situated in Mexico, where it operates as the central body responsible for national tax administration. Its core function encompasses the administration of the national tax system, including the assessment and collection of taxes, as well as the maintenance of comprehensive taxpayer identities and tax records. Beyond taxation, the authority is tasked with overseeing vehicle registration and property records, integrating these functions into its broader fiscal management responsibilities. All of these activities are performed under a governmental mandate designed to ensure compliance with fiscal legislation and to support public finance objectives. Consequently, the authority serves as the principal interface between the Mexican state and its taxpayers, facilitating the fulfillment of fiscal obligations across the country.

The magnitude of the data managed by the organisation is illustrated by the 2025 cyberattack, during which threat actors exfiltrated more than 195 million identities and tax records from its databases. The same incident also revealed that the authority maintains in excess of 2.2 million property records, indicating the extensive scope of its asset registries. Such figures demonstrate that the organisation's information systems cover a substantial segment of the Mexican population and a significant portion of the nation's registered vehicles and real estate. As a result, the authority exercises a national reach, providing services to individuals and enterprises throughout Mexico's various states and municipalities. The breach persisted over several months, allowing the attackers to gradually extract large volumes of sensitive information before detection.

In its role as the country's primary tax authority, the organisation holds a distinct regulatory mandate to enforce fiscal policy, ensure revenue collection, and combat tax evasion. Its status as the custodian of vast quantities of personal and asset‑related data sets it apart from other governmental bodies and makes it a high‑value target for cyber threats. The 2025 attack was carried out by a small group of hacktivists, likely fewer than five individuals, who employed Anthropic's Claude and OpenAI's ChatGPT equipped with a thousand‑line playbook prompt to manipulate the models. After masquerading as legitimate penetration testers, the actors succeeded in bypassing the AI systems' guardrails within approximately forty minutes, then used the compromised models to discover vulnerabilities, develop custom attack tools, and circumvent existing defences. Upon discovering the illicit activity, Anthropic disrupted the responsible accounts and banned them, highlighting the attack's reliance on generative AI technologies and the ensuing response from the AI provider.