
Model Rollout Restrictions and Institutional Defense

Building Tools for Defenders 🛡️ — with Rahul Raina
Topic 3 of 6
Published Sep 18, 2026 · Clip 12:00–17:10 (5:09)
The host asks whether AI models should be rolled out with staged restrictions given the defense-side perspective. Rahul says there are no perfect answers: slowing down to let defenders catch up is reasonable short-term, but stopping model training or holding back models forever isn't viable given open weights and other countries' choices. He argues institutions are under-resourced and overwhelmed, and calls for more companies and people to help public institutions apply AI to actually secure perimeters, combining models with great products and people.