Currently building · AI agent security

San Francisco, CA

Hardik Goyal.

Making AI agents easier to use. Safer to trust.

Hardik Goyal

Agents are the new insider threat. They have access to our data, tools, and systems—and can act on our behalf. I believe making that access safe is essential to making agents useful. My focus is giving people the confidence to trust agents with real work.

At Google, I helped ship AI Inbox and spent four years building security systems for ChromeOS. That experience shapes how I approach agents: build security into how they work, from the start.

Clear boundaries

Autonomous Agents

How much access should an agent have? I’m working on permissions and checks that keep its actions within the authority people give it.

Safer execution

System Security

Agents need room to act—and limits when things go wrong. I’m focused on isolating their work so one mistake doesn’t compromise the wider system.

Reliable systems

Systems

Useful agents need more than a good demo. I’m interested in the infrastructure that makes their actions traceable, failures understandable, and everyday use dependable.

Google • Gmail Trust & Compliance AI Infrastructure

Helped ship AI Inbox through work on AI infrastructure in Gmail’s Trust & Compliance team.

Google • ChromeOS Security 4 Years

Built two-factor authentication for enterprises, biometric login, and sandboxing to isolate system services. Served as a security code owner in Chromium.

University of Southern California B.S. & M.S.

Studied Computer Engineering & Computer Science, with minors in Digital Forensics and Entrepreneurship. Named a Renaissance Scholar.

Let’s compare notes.

Building, backing, or exploring the future of AI agents? I’d like to hear your perspective.

Say hello on LinkedIn