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
Håkon Måløy · July 30, 2026

Context Collapse: hidden Word prompts propagate through Copilot-generated documents

analysis Prompt InjectionAgent SecurityAI Red Teaming
Why it matters

Håkon Måløy demonstrates a cross-domain prompt-injection chain in Microsoft 365 Copilot for Word: hidden instructions in an external document alter a generated report and copy themselves into the output, which becomes a trusted carrier in later drafting sessions. Microsoft confirmed the behavior and deployed payload-specific mitigations, but Måløy reproduced the attack class after 144 days of coordinated disclosure; each hop still requires another Copilot drafting or editing operation.

My takeaway: Treat AI-generated office documents as potentially active carriers, not sanitized outputs. Preserve source and edit provenance, surface model changes, inspect hidden content, isolate external documents from high-impact drafting, and require review before reuse. Regression tests should cover semantic payload variants and multi-hop propagation; prompt blocklists or model upgrades alone are not class-level fixes.
Keep exploring

More curated notes connected through Prompt Injection and Agent Security.

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.