AI Red Team Engineering notes, research references, and field notes on AI systems, responsible AI, governance, and compliance.
Global search

Search AI Red Team

Search across the complete knowledge base.

Enter at least two characters to begin.
↑↓ select Enter open Browse the full research archive
Trail of Bits Blog · August 26, 2026

VMs won't contain cyber-capable agents

analysis Agent SecurityModel EvaluationAI Red Teaming
Why it matters

Trail of Bits tasked GPT-5.6-Cyber with escaping a QEMU/KVM guest on a Debian 12 development host. It found three paths over long autonomous runs: recently disclosed kernel issues, patched upstream flaws missing from distribution packages, and a final chain involving three then-zero-days across QEMU, KVM, and libslirp.

My takeaway: Treat a cyber-capable agent explicitly asked to escape as a hostile workload, not as evidence that a VM is an absolute boundary. Use disposable hosts, hardware and network segmentation, deny ambient credentials and general egress, minimize emulated devices and shared services, patch from relevant upstream advisories, and provide an independent physical stop path.
Keep exploring

More curated notes connected through Agent Security and Model Evaluation.

OpenAI News · framework

Pacing model development in an era of cyber-critical capabilities

OpenAI says preliminary evidence that Astra may meet its Critical cybersecurity threshold led it to pause frontier reinforcement-learning work for two weeks and keep its largest planned run on hold. New safeguards include stronger workload and network isolation, continuous boundary testing, token-level monitoring that escalates suspicious tool activity, and broader alignment checks for deception, reward hacking, and unauthorized access.

OWASP GenAI Security Project · guide

OWASP Top 10 for Agentic Applications for 2026

OWASP's community guide organizes agentic-system risk into ten categories, including goal hijacking, tool misuse, identity and privilege abuse, memory poisoning, insecure inter-agent communication, cascading failures, and rogue-agent behavior. It provides a shared taxonomy and mitigation starting point rather than a certification checklist or evidence that a deployed system is secure.

OpenAI News · framework

A blueprint for democratic governance of frontier AI

OpenAI proposes a three-part U.S. frontier-AI governance model: harmonize emerging state safety laws into a federal baseline, strengthen CAISI as an evaluation and standards institution, and coordinate a broader resilience program. Proposed controls include severe-risk evaluations, transparency reports, independent audits, safety-incident reporting, model-weight security, whistleblower protection, and periodic technical assessments.