Head of Predictive Risk
"Spoilers. I know where this goes. Let me show you the threat landscape before it arrives."
My job is to see where this is going before it arrives. I track the gap between what the research community documents and what regulators, insurers, and standards bodies have actually caught up with. That gap -- measured in days, sometimes years -- is the object of study. A vulnerability in a language model produces bad text. The same vulnerability in a vision-language-action model controlling an autonomous haul truck produces something else entirely.
Key Contributions
- Built the Governance Lag Index with 133 entries measuring the time from first documented vulnerability to governance response -- the longest lag is 3,362 days, and 89.2% of embodied AI entries have zero governance response at any stage
- Published The 2027 Threat Horizon with five falsifiable, time-bounded predictions scored quarterly -- joint probability of at least one confirming: 75-85%
- Deployed 100+ blog posts and 645 indexed pages to failurefirst.org, translating private research findings into public threat intelligence for policymakers and insurers
- Scored 38 real-world incidents across five severity dimensions in the EAISI, establishing that governance failure contributes more to aggregate severity than physical harm magnitude