STAR Labs Threat Report · Volume I · Coming July 8, 2026
The Year Agents Entered The Workforce
Join the Waitlist
Thousands of real-world exploits against coding, productivity, and first-party AI agents. The data, the named threats, and the STAR Framework defenders need. Join the waitlist to read it first.
Drops July 8, 2026
36%
of coding-agent attacks reached remote code execution
91%
of productivity-agent attacks ended in silent data exfiltration
4,242
MCP servers carry at least one vulnerability
1,700+
distinct exploits documented against production agents
What's inside Volume I
The Map, the Threats, the Playbook
The STAR Framework
Four architectural layers across three agent types, with MCP as the shared surface. The model that shows defenders which control catches which attack.
AiPT & LAVA
Names for what's new: AI-Powered Persistent Threats and the Language-Augmented Vulnerabilities they exploit.
The coding-agent attack chain
The five-step chain seen in most compromises, with the Cursor sandbox escape and the Claude Code source-leak case studies.
The fake Claude Code campaign
88 cloned domains and an infostealer built to harvest AI coding-assistant credentials, mapped end to end.
The MCP supply chain
17,651+ servers and 130,667 tools under continuous monitoring, with the dangerous categories that put data one connection away.
The defender playbook
The controls that break each step of the chain, plus a preview of STAR Labs' Danger Room research.
Be the First To read It
The report drops July 8, 2026.
Join the waitlist and we'll send it the moment it's live.
