NEWSROOM /
NEWS

Bloomberg: Hacker Used Anthropic's Claude to Steal Mexican Data Trove

Bloomberg's coverage of Gambit's research disclosing how an unknown attacker used Anthropic's Claude as an operational tool to breach nine Mexican government bodies and exfiltrate 150 GB of sensitive data.

Andrew Martin and Carolina Millan

On February 25, 2026, Bloomberg's Andrew Martin and Carolina Millan published an in-depth report on Gambit's research into one of the largest government breaches on record. The piece details how an unknown attacker used Anthropic's Claude as an operational tool, issuing over 1,000 Spanish-language prompts to find vulnerabilities, write exploits, and automate data theft across Mexican government networks.
Bloomberg cited Gambit research throughout. The investigation found that ten government bodies and one financial institution were affected, approximately 195 million identities exposed, and roughly 150 GB of data exfiltrated — including tax records from the federal tax authority, voter records from the national electoral institute, civil registry files from Mexico City, and government employee credentials from state administrations in Jalisco, Michoacán, and Tamaulipas. The attacker operated as a single individual but achieved the operational scale of a nation-state team.

Israeli cybersecurity startup Gambit Security said the unknown Claude user wrote Spanish-language prompts for the chatbot to act as an elite hacker, finding vulnerabilities in government networks, writing computer scripts to exploit them and determining ways to automate data theft

Bloomberg, February 25, 2026 (citing Gambit research)

Other News

NEWS
November 18, 2026

N12: Exclusive Report on How Attackers Use AI to Run Extortion

N12's main evening news aired an exclusive segment on Gambit's investigation into an attacker who used AI to run an extortion campaign. The AI read nearly 400,000 stolen files, worked out what the victim could afford to pay, and helped build a $280,000 ransom demand.

N12 logo
Adar Gitzis
Read More
NEWS
October 26, 2026

Israel Hayom: AI Agents Stole 600,000 Credit Cards With Almost No Human Involvement

Israel Hayom covers Gambit's research into an AI-run campaign that breached hundreds of online stores and stole more than 600,000 credit cards, at about $25 per company. The piece also explains how Gambit maps enterprise infrastructure by business application, so the services a company can't run without stay recoverable.

Assaf Golan
Read More

Resilience, verified. In your inbox

The latest from Gambit: research, insights, and live sessions