Point of view · Jul 2026
You Made an AI Policy. So Now What?
A policy is not governance. The five ways an AI council fails, mapped to Lencioni's five dysfunctions, and the cheapest fix available before its next meeting.
Read →Media
Blog
Point of view · Jun 2026
From Chaos to Control: Managing AI Risk in Healthcare
A control framework tells you what to do. Whether a healthcare organization can actually do it is a question of people, ownership, and maturity.
Read →
Point of view · Jun 2026
You Can't Patch AI!
Policy borrows patch-management language for AI systems. But fine-tuning, retraining, and weight editing are not patches, and the difference matters.
Read →
Research · May 2026
Your Temperature Setting Is a Bigger Security Risk Than Your Quantization Tier
Our SPIE DS26 paper ran 14,850 evaluations across five model families and three quantization tiers. Temperature is the safety variable to watch.
Read →
Point of view · Feb 2026
Why Anthropic's Capability Leak is an AI Maturity Problem
Chinese labs drained Claude's intelligence with 16 million automated exchanges. The AI industry has an operational maturity problem, not just a security one.
Read →
Field notes · Jan 2026
2025: The Year of the Spark
A look back at our first year: research at the NATO IST-210 symposium, CAMLIS, BSides Las Vegas and SPIE DCS, and where we take it next.
Read →
Research · Nov 2025
Beyond the Firewall: The Three Pillars of Offensive AI Security
AI fails in ways firewalls cannot see: data poisoning, model theft, prompt injection. The three layers of a deliberate offensive AI security strategy.
Read →
Point of view · Nov 2025
Fast Code, Soft Skills?
Developers once fought over whether IDEs were cheating. Now it is AI assistants. What automation research says about losing the fundamentals.
Read →
Research · Jun 2025
Adversarial Threat Vectors and Risk Mitigation for RAG
Our research on how attackers exploit every stage of a RAG pipeline, from poisoned vector stores to leaked prompts, and the controls that stop them.
Read →
Point of view · Apr 2025
Is AI Security a Distraction?
The industry is racing to bolt security onto AI systems that were never designed to be safe. Securing an unsafe system is the wrong place to start.
Read →
Research · Apr 2025
AI Safety Scores
An independent Future of Life Institute assessment graded leading AI companies on safety, and no vendor scored above a C. The vendor's risk becomes yours.
Read →