The Next AI Wave Requires Stronger Cyber Defenses, Data Management

Federal IT leaders are prioritizing vulnerability assessments, data management and workforce training as AI becomes commonplace in government, officials explained during GovCIO Media & Research’s 2025 Federal IT Efficiency Summit in Tysons Corner, Virginia, last week.

“With AI being embedded in so many of our tools and technologies, we are already finding that agencies are not able to really provide the additional protections for their data and our personally identifiable information,” Jennifer Franks, director of the Center for Enhanced Cybersecurity and acting director of the Analytics Foundry at the Government Accountability Office (GAO) said. “Going into the next phase of deploying and developing additional AI software tools and technologies, we’re finding more and more that federal agencies do not have the capabilities to really put us at the edge of where it is we need to go.”

Franks said her team at GAO is doing an AI tech assessment, working with Congress to evaluate 16 critical infrastructure sectors for vulnerabilities over the next few months.

“A transportation sector is going to have different metrics and policy measures than an energy sector. However, they are going to intersect at some point based on the customers and the personally identifiable information that is built into those large language models,” Franks told the audience.

AI Helps Link Private Sector to Government

Agencies are managing emerging AI risks while harnessing its potential to enhance operations. Cheryl Ingstad, deputy director of the Digital Platforms and Developer Ecosystem at the Defense Innovation Unit (DIU) outlined how the Digital OnRamp platform leverages AI to link the private sector with the government on emerging technologies.

“The problem we’re trying to solve here is that small companies, medium-sized large companies, academia have tremendous technology that we need in the Defense Department, but they don’t know where to find their DOD customer,” Ingstad said. “We’re providing a digital platform that offers security.”

Ingstad said Digital OnRamp will start “at least at IL2 level,” allowing the use of large language models and generative AI tools to search DOD databases to find partnership opportunities.

CEO of Knox Systems Irina Denisenko said that AI tools have allowed smaller companies to participate in government, driving efficiency, innovation and competition.

“What we’ve seen particularly over the last few months is in many avenues, smaller companies, more nimble companies, but companies that have very powerful technology that can unlock so many critical missions are able to participate in not just acquisition, not just on the cyber front, but also on the policy front,” Denisenko said.

Streamlining Data to Maximize AI’s Potential

While AI tools have helped sort and sift through millions of data points in maritime domains, the Naval Research Laboratory (NRL) has faced challenges in uniting all of the data points under one agency and workforce that is capable of leveraging the data to the best of their ability, Head of NRL’s Mission Development Branch Alan Hope explained.

“We have a very diverse user base of different levels of capability of users. So you need to build systems that are easy to use yet still have the smarts behind them for those super users to go in and exercise all the functions and analytics to do the deep work that needs to be done to solve the really hard problems,” Hope said.

Hope suggested that “consolidating commercial data buys and a common system for doing ship tracking” could be a huge step toward efficiency that “could save money, time, space, storage requirements, processing requirements and allow others to build applications for analytics within that space.”

Doyle Choi, senior director of strategic federal programs at Pure Storage, added that as AI technology becomes more ubiquitous, there will be an important role to play in shrinking the physical and environmental footprint of AI storage.

“Imagine the efficiencies achieved by collapsing 70 racks of storage down to 10, reducing power consumption from over 400 kilowatts down to 60, and simplifying their data management through a common user experience,” Choi said during the summit.