Expert Insight: Dr. Chenxi Wang on the Future of AI and Cybersecurity
By Evan Lybrand
Media InquiriesThe Information Networking Institute’s Future of Secure AI executive certificate is designed for executives and aspiring leaders who are navigating the opportunities and security risks posed by rapidly evolving Artificial Intelligence (AI). Through this in-person certificate program, expert instructors share their insights and actionable strategies on implementing, securing and scaling new AI capabilities.
One featured guest instructor is Dr. Chenxi Wang, founder and managing general partner at Rain Capital, a venture firm based in Silicon Valley. Her session, “The Future of AI and Cyber: Landscape, Threats and Opportunities,” will examine current trends in how cybersecurity interacts with AI. We interviewed Dr. Wang to understand her thoughts on the field and what participants can expect from her session.
In your work at Rain, you’re connecting with founders who are helping shape the infrastructure of AI. What trends are you noticing?
Dr. Wang: AI as a field has progressed very quickly, with innovations coming fast and furious. What we have seen in the last three months is the concept of recursive, self-improving agents. This type of agent is all the rage right now in Silicon Valley. It’s essentially infrastructure that supports a self-improvement training loop, where an agent and its improving mechanism work together. It’s a really interesting way to think about how autonomous agents are evolving.
There's been a lot of talk recently about the need for an AI slowdown. What do you make of this debate?
Dr. Wang: I think some of it is posturing. One way to look at it is, “We are here already, so we don’t want anybody else to get here.” You can look at it that way, or you can think of it as an interesting bump in the road when a piece of technology is moving too fast.
Certainly, there are risks to adopting AI for critical applications and critical decision-making. However, I think the right thing to do is not slow down the innovation because I don’t think you can actually slow it down. Even if Anthropic slows down, others will continue to invest. I think the right thing to do is to make sure the innovation is done in a way that is safe and trustworthy. Because of that, the security and safety measures have to come with innovation, rather than being bolted on at a later time.
In the Future of Secure AI executive certificate program this November, you’re going to be leading a session on the future of AI and cyber. Can you give us a teaser on what that session will cover?
Dr. Wang: We are looking at autonomous penetration testing being the table stake, because Large Language Models (LMs) have been shown to be really good at finding vulnerabilities at scale and constructing exploits. They can remediate some of it, but what they have not yet been demonstrated to do effectively is defense.
Attacking and defending are different. As an attacker, you only have to find one way in, and you don't care how it impacts the rest of the system. But as a defender, you have to make sure the system is up and running and that the application is serving its users as well as defending against attacks. The job on the defense side is much harder.
What that means is we need to work on AI models for defense that are more robust and that don’t just care about achieving a single goal — they have to achieve a multitude of goals at the same time. It will require a completely different decision-making paradigm. In that session, I’ll also talk about how AI technology can be adapted for defense rather than just offensive tasks.
When you look at how AI innovation and cybersecurity intertwine, what fascinates you the most and what worries you?
Dr. Wang: That’s an interesting question. I think for security with AI, most of the traditional security stack — all the way from application security to network security and infrastructure security — has to be rewritten because the risks have changed and the compute infrastructure has changed with it. So, we need brand new layers in the security stack. That’s a big opportunity, and that excites me.
What worries me is that AI is a powerful tool, a powerful tool for attackers as well as for defenders. And the defense side is still a little bit lacking today. So, it’s likely we’ll see some large breach that’s coming our way if we’re not careful.
Interested in joining the Future of Secure AI program? are due Friday, October 23, 2026, and spots are limited. Reach out to ini-execed@andrew.cmu.edu with any questions. If you aren't able to participate this November, to be notified of future programs.