Why NVIDIA Open Sourced Its Agent Security Layer | Lightwork

NVIDIA just open sourced the piece it says was missing from the agent stack. Claire Zau talks with Ali Golshan, Senior Director of AI Software at NVIDIA, about OpenShell and Sentry, and why a chip company is giving the runtime away under Apache 2.0 with more than a hundred partners already on board. His argument is that we have been trying to secure agents inside an environment never built for them, and the fix is not a better prompt or one more guardrail in the harness. He believes enforcement has to sit outside the agent's own process, where it cannot reason about it or route around it, while still leaving the agent room to work. He walks through what the recent agent breakouts actually had in common, and why none of them surprised him. He also explains what opened up once teams could trust fleets of agents on long running tasks, why he keeps coming back to SSL and the early web, and the gaps he wants founders building in, from agent-native identity to policy verification. Chapters: 0:00 OpenShell and Sentry Explained 2:42 Why Make OpenShell Open Source? 4:41 What Can AI Trust Unlock? 7:07 Why Companies Are Joining OpenShell 8:55 Rethinking Sandboxing for AI Agents 11:28 Rethinking AI Agent Safety 13:36 What Some People Are Missing in AI Security Stay in touch: www.lsvp.com X: https://twitter.com/lightspeedvp LinkedIn: https://www.linkedin.com/company/lightspeed-venture-partners/ Instagram: https://www.instagram.com/lightspeedventurepartners/ The content here does not constitute tax, legal, business or investment advice or an offer to provide such advice, should not be construed as advocating the purchase or sale of any security or investment or a recommendation of any company, and is not an offer, or solicitation of an offer, for the purchase or sale of any security or investment product. The views expressed by our guests do not necessarily represent the views of Lightspeed. For more details please see lsvp.com/legal.