Registration for CTRL+AI 2027 is now open →
.svg.webp){kind=logo}
Tackling the hardest questions in AI security, privacy, governance, and compliance.
Where strategies are tested, assumptions are challenged, and the path to scalable, trustworthy AI becomes clear.
Registration + Networking
Grab your badge, a coffee, and a bite to eat. This is your time to connect with fellow attendees before the program begins.
Welcome + Keynote
Virtue AI CEO and Co-Founder Bo Li opens the day with a special keynote message from Intel CEO Lip-Bu Tan.
In Conversation: Stress-Testing Agentic Security
Agentic systems don't fail the way traditional software fails... and most security teams aren't ready for that. In this discussion, Sandeep Pombra (NVIDIA) gets specific: How do you stress-test an AI system that reasons and acts autonomously? What does a real test plan look like? Afterwards, Wenbo Guo (UC Santa Barbara / Virtue AI) will join Sandeep for an expanded conversation and Q&A.
Virtue AI Research: Building Safe + Secure Agentic AI
The stakes around agent safety and security have never been higher. In this presentation, Dawn Song (UC Berkeley / Virtue AI) shares foundational research exploring the challenges and opportunities for agentic systems in the real world, including demos of PolicyGuard and AgentSuite.
Lunch
Eat, mingle, and recharge.
AI Enforcement Has Arrived:
Securing + Governing AI in 2026
This panel cuts through the noise to examine what's actually happening on the frontlines of AI security and governance. Our panelists will unpack compliance priorities, organizational blind spots, and the gap between written AI policy and real-world enforcement.
AI Risk in the Enterprise: Data Security + AI Trust
Every enterprise AI initiative sits on a foundation of sensitive data. That foundation is also its most exploitable attack surface. This panel brings together enterprise leaders to examine the real tensions between AI acceleration and responsible risk management, and what it actually takes to build AI systems enterprises can trust.
Virtue AI Research:
Safety + Security in an Age of Self-Improving AI
Carlos Guestrin and Sanmi Koyejo examine why AI evaluation methods systematically mislead us and what rigorous, policy-grounded measurement actually requires. They will also confront what self-improving AI means for the security surface and make the case for safety and security infrastructure that keeps pace with systems that never stop changing.
Happy Hour
We have the brightest minds in AI and AI security, all in one room. Have a drink and say hello!
San Francisco's most iconic location. AI's most important conversation.
Building 103, 103 Montgomery St, San Francisco, CA 94129
Arrive early to connect with peers and enjoy.
Your questions, answered.
A one-day conference on AI security, governance, trust, and risk. Designed for people who are committed to solving the problem of secure, scalable AI.
Executives, researchers, practitioners, and operators who are making real decisions about how AI gets built, deployed, and governed. If AI risk, compliance, or security is part of your job, you belong here.
This is an exceptionally rare gathering: researchers at the frontier of AI security, executives accountable for governance at scale, and operators who've seen what happens when things go wrong.
CTRL+AI is presented by Virtue AI, the leader in AI security, governance, and compliance. We didn't build this conference to fill a calendar. We built it because these conversations need to happen.
The conference will take place at The House, nestled within the historic Presidio in San Francisco.
Paid parking is available nearby, rideshare drops off directly at the venue, and full directions will be in your confirmation email.
Research Passes are available to students and researchers by application.
It's selective, not secretive.
Industry Passes are available to qualified practitioners and executives.
Don't let safety be the friction that grounds your innovation. Virtue AI is the launchpad for agentic systems, clearing the path from research to production-ready performance.
.png)
.jpeg)
