Indiana CIO Warren Lenard on AI in State Government and the AI vs AI Battlefield

Indiana State CIO Warren Lenard and Nathan Kirk on NR Labs Podcast Episode 4: A Battlefield of AI Versus AI

NR Labs Podcast, Episode 4 · Host: Nathan Kirk · Watch on YouTube

On Episode 4 of the NR Labs podcast, our cybersecurity podcast for technology and GRC leaders, host Nathan Kirk sits down with Warren Lenard, Indiana CIO and head of the Indiana Office of Technology. They cover Indiana's statewide Microsoft Copilot rollout, why AI output always needs a human check, the risks of AI-generated code, data classification and retention, and what Warren calls "a battlefield of AI versus AI."

Watch and follow

Who is Warren Lenard?

Warren Lenard is the State Chief Information Officer of Indiana and leads the Indiana Office of Technology (IOT), which delivers shared technology services to every agency in Indiana state government. He took the role in 2025 after about a decade leading technology in retail and consumer businesses. He is also a leader in SIM Indy, the Indianapolis chapter of the Society for Information Management.

How is Indiana using AI in state government?

Indiana ran an eight-week Microsoft Copilot pilot with about 200 users, analyzed the results for several months, then rolled out Copilot Chat followed by full Copilot licenses. Every state employee now has access. Indiana's Management Performance Hub sets AI policy, while the Indiana Office of Technology deploys the tools and runs an AI center of excellence.

Warren's starting point was the data. "Your AI deployment is only going to be as good as your data is," he said. The pilot surfaced issues the team hadn't known about, which made data classification a priority. Warren called it "an eye-opening pilot."

What is AI vs AI in cybersecurity?

AI vs AI describes a threat landscape where attackers and defenders are both using AI, so defensive AI tools are up against AI-powered attacks. "We are slowly or quickly getting to a battlefield of AI versus AI," Warren said. Deepfakes are a big part of it, and he expects them to get good enough that people can't spot them alone: "You're going to need tools that help you and say, this is not real."

Building defenses for that kind of threat is the focus of our Cyber Innovation services.

Should you verify AI outputs?

Yes, always. Warren's approach to AI hallucinations is "education, education and control," the same way organizations train people to spot phishing. Treat anything AI produces as a framework: take the speed, then verify every claim you can't confirm yourself.

"It's a powerful, powerful tool," Warren said, but even for a simple email, "please read it through and make sure it says what you want it to say."

Is AI-generated code secure?

Not without review. AI can "shave weeks, maybe even months off of a software development project," Warren said, but the code still needs security review, documentation, and comments, the same disciplines experienced engineers have always applied.

His bigger concern is citizen developers. When people with no coding background build applications with AI, Warren warned it "can create an entire world of applications built in the corners of organizations that people are not aware of till it's too late." That's a new wave of shadow IT, and those apps often lose their owner when the builder moves on.

How does data classification differ between the public and private sector?

The rules aren't really different, because the risk is the same. As Warren put it, a breach or leak of PII or PHI is a headline that "is not good for anybody." A government agency won't go out of business the way a company can, but it still faces the fallout.

For defense contractors handling controlled unclassified information, the same classification work is the foundation of CMMC compliance.

How long should organizations keep data?

There's no single answer, and that's the problem. Warren said every organization he's worked in has struggled to agree on a data retention policy: "We need this data forever, but we don't want this data forever, because this data might hurt us if we have it forever." Sometimes legal wants to keep everything, and sometimes IT wants to purge 20 years of records that slow systems down.

The answer comes from balancing what you must keep, what you'd like to keep, and who has access. "There's nothing sexy really about that, but it needs to be done," Warren said, so that AI tools only touch the data they should. Our Advanced GRC services help organizations build classification and retention programs that hold up.

Key takeaways

  • Classify and govern your data before deploying AI.
  • Run a pilot, and treat the problems it surfaces as wins.
  • Verify AI output like a first draft.
  • Hold AI-generated code to normal engineering standards.
  • Prepare for AI-powered attacks and deepfakes now.

Watch the full episode

Hear the full conversation between Nathan Kirk and Warren Lenard: Watch Episode 4 on YouTube. For more GRC podcast and cybersecurity conversations, subscribe to the NR Labs YouTube channel and follow NR Labs on LinkedIn.

FAQ

Does Indiana use Microsoft Copilot?
Yes. After an eight-week pilot with about 200 users, Indiana rolled out Copilot Chat and full Copilot licenses, and every state employee now has access.

What is AI vs AI in cybersecurity?
It's a threat landscape where attackers and defenders both use AI, including AI-generated malware and deepfakes, so defenders need AI-powered tools to keep up.

Should you verify AI outputs?
Yes. Warren Lenard recommends treating AI output as a framework and verifying any claim you can't confirm yourself, because AI can hallucinate convincing but false information.