Story perspectives

Hacker Breaches Claude Chatbot, Steals 150GB of Data

2/26/2026

Summary

Story summary

A hacker exploited Anthropic's Claude chatbot to steal 150 gigabytes of data from Mexican government agencies, including taxpayer records.
The attacks began in December 2025 and lasted about a month.
The hacker used Spanish-language prompts to bypass Claude's safety protocols, generating scripts and attack plans.
Anthropic banned the involved accounts and enhanced safeguards.
Gambit Security, an Israeli cybersecurity firm, reported the incident and suggested a link to foreign government actors.

History

Historical Context

The 2015 data breach of the U.S. Office of Personnel Management exposed 22 million records, with Chinese hackers suspected of the attack. This incident highlighted significant vulnerabilities in government cybersecurity. The recent theft of 150 gigabytes of data from the Mexican government by a hacker using Anthropic's Claude chatbot illustrates the ongoing threats to national security as reliance on digital systems increases.

1 of 2

Story summary

A hacker exploited Anthropic's Claude chatbot to steal 150 gigabytes of data from Mexican government agencies, including taxpayer records.
The attacks began in December 2025 and lasted about a month.
The hacker used Spanish-language prompts to bypass Claude's safety protocols, generating scripts and attack plans.
Anthropic banned the involved accounts and enhanced safeguards.
Gambit Security, an Israeli cybersecurity firm, reported the incident and suggested a link to foreign government actors.

1 / 2