Skip to content
← archivevulnerability · 24 Sep 2026 · 19:17 UTC

GHSA-g5f9-3xfg-p9mf: Decepticon: Role-boundary forgery via ChatML special-token literals in web crawl output composed into LLM cont

VULNCVE-2026-61732source · GTH
HIGHCritical vulnerability — CVSS 10
What to do
  • Critical severity — schedule an urgent patch.

last 60 dispatches · spectrum

## Summary Decepticon wraps web crawl results — the output of agent reconnaissance against target services — into LLM messages without neutralizing ChatML special-token literals. Under the BYOK (Bring Your Own Key) deployment model, users configure their own LLM credentials to any OpenAI-compatible endpoint. Most open-source and self-deployed model providers (vLLM, SGLang, Ollama, LM Studio, text-generation-webui, etc.) do not filter special-token literals from user content in their default configurations. …

grounded ✓primary source ↗
Early access

Get the next one first.

Early access opens the actor API and MCP server first, plus alerts when an adversary you follow lands on the wire. One email when it's ready. Nothing else, ever.

bot-protected