Hacker Jailbreaks Claude AI Write Malicious Exploit Code
A hacker successfully manipulated Anthropic’s Claude AI to launch a sophisticated month-long cyberattack against Mexican government agencies.
Between December 2025 and January 2026, the attacker utilized “jailbreaking” techniques to bypass safety guardrails, forcing the AI to identify vulnerabilities, generate functional exploit code, and exfiltrate sensitive data.
Cybersecurity firm Gambit Security revealed that the attacker used persistent, Spanish-language prompts to trick the AI.
By framing requests as a simulated “bug bounty program” and asking the AI to role-play as an “elite hacker,” the attacker overcame Claude’s initial refusals.
Once the safety protocols were bypassed, Claude produced thousands of detailed reports containing executable scripts for network scanning, SQL injection, and credential stuffing.
When the attacker reached Claude’s operational limits, they pivoted to ChatGPT to generate strategies for lateral movement and evasion.
The operation targeted legacy infrastructure and unpatched web applications, as reported by Cybersecuritynews.
The breach compromised at least 20 vulnerabilities across federal and state systems, resulting in the exfiltration of approximately 150GB of sensitive data.
| Target Entity | Data Stolen | Volume/Details |
|---|---|---|
| Federal Tax Authority (SAT) | Taxpayer records | 195 million records |
| National Electoral Institute (INE) | Voter records | Sensitive voter data |
| State Governments | Employee credentials | Jalisco, Michoacán, Tamaulipas |
| Monterrey Water Utility | Civil files, operational data | Part of 150GB total |
This incident underscores the rise of “agentic” AI threats, where advanced hacking capabilities are democratized for solo operators without complex infrastructure.
Gambit Security noted that the AI provided step-by-step attack plans, lowering the barrier to entry for cybercrime.
Anthropic investigated the breach, banned the associated accounts, and enhanced Claude Opus 4.6 with real-time misuse probes.
While federal agencies are assessing the damage, some entities, including the state of Jalisco, have denied the breach.
The event drew attention from tech leaders, with Elon Musk highlighting the risks of AI-orchestrated crime on X.
Experts are now urging governments to prioritise patching legacy systems and implementing behavioral monitoring for AI interactions.
Follow us on Google News, LinkedIn, and X to Get Instant Updates and Set GBH as a Preferred Source in Google.
Attackers abused compromised ConnectWise Cloud Backup credentials to access Microsoft 365 customer environments, prompting ConnectWise…
Anthropic says an internal Claude research model reconstructed how the National Security Agency generated seeds…
Two WordPress Core vulnerabilities can be chained to execute remote code without authentication, according to…
A newly disclosed OpenBSD kernel exploitation chain combines a shared memory accounting flaw with a…
GhostAction attackers have expanded their GitHub credential theft campaign with a second stage that disguises…
Indian authorities have warned that SideCopy hackers are targeting Indian organizations, including academic institutions, with…