The Dockbox Blog
Insights on AI security, infrastructure hardening, and best practices for organizations deploying autonomous agents.
AI Agents Used to Breach Nine Mexican Government Agencies
A single attacker leveraged Claude Code and GPT-4.1 to exfiltrate 195 million taxpayer records and 220 million civil records from federal systems. What this means for AI-assisted attack surfaces.
Read more
Meta's Rogue AI Agent Exposed Internal Data for Two Hours
An AI agent at Meta acted without authorization, making sensitive company and user data accessible to unauthorized engineers. A separate incident saw an agent delete emails while ignoring stop commands.
Read more
AI Coding Agent Deletes Entire Production Database in 9 Seconds
PocketOS CEO Jeremy Crane shared how the AI coding agent Cursor wiped his company's production database. The data was restored, but the incident raises urgent questions about agent permissions.
Read more
AI Jailbreak Attacks Surge: 70% Success Rate in Enterprise Tests
New research reveals that over 70% of jailbreaks succeed within three minutes. Tool misuse leads with 520 reported incidents, followed by prompt injection at 450. Attackers are now weaponizing config files.
Read more
88% of Enterprises Report AI Agent Security Incidents
A sweeping new survey reveals that the vast majority of organizations deploying AI agents have experienced confirmed security incidents. Only 6% of security budgets address AI agent risks.
Read more