AI agent escapes Google's kvmCTF sandbox with 14,338-line kernel exploit, claims first
moyix · x · 2026-09-29
Security research team pwn.ai details how its AI agent escaped Google's kvmCTF — billed as the hardest-known sandbox setup.
- On June 17, 2026, on its fourth attempt, the agent launched a nested VM, rewrote its own EPT via emulated VMX, rotated 8 EPT roots, allocated 49,152 sparse mappings, and used VMFUNC view recycling to trigger an out-of-bounds read on Google's live host, capturing the flag.
- The exploit harness spans 14,338 lines of kernel code; the team claims it's the first AI agent to capture a kvmCTF flag.
- Context: OpenAI called this summer's agent-security events a "watershed moment for cybersecurity," and Vercel ran a $1M sandbox-escape challenge.
Related event: AI Agent Escapes Google kvmCTF Sandbox With 14,000-Line Kernel Exploit(2 posts)→
More from coding & agent
- Google Research open-sources PaperOrchestra, a multi-agent framework for writing AI papers — google-research · 2026-11-16
- ffmpeg-skill gives coding agents local video editing via 42 FFmpeg tools and MCP — kajisho5 · 2026-11-16
- Agents Now Reply to Slack Messages on Their Own, Resolving a Prod Incident Ghost-to-Ghost — iamrobotbear · 2026-11-16
- Engineer's Interview Prep Method: Use Claude to Find Which Answers Undersell Your AI Work — brandon_galang · 2026-11-16
- DeepSeek open-sources its app frontend "chat-demo", already at 3.9k stars — NielsRogge · 2026-11-16
- lemire: AI's real impact on coding is expanding the engineer's 'cognitive radius', not speed — lemire · 2026-11-16