Kansas CISO John Godfrey said machine learning traditionally has been helpful with tasks like log management and malicious email scanning and blocking. But recently, uses for the technology “are starting to ratchet up a notch,” he added.
One emerging idea is to use AI for “patch management at scale.” Updates need to happen regularly, and ideally, cyber teams would deploy software patches to all machines the moment the patches become available. But this isn’t always possible: Teams must also be careful to ensure patching doesn’t disrupt operations, and it’s customary to wait a few weeks before applying a patch. AI might help balance the needs for speed and stability, however. The technology could help assess the likelihood of patching going smoothly.
“What data can we capture on the success rate of patches, and the extent to which those are disruptive to the systems, and how can that inform our patching strategy going forward?” Godfrey said. “If we know that 90 percent of the time these patches are not disruptive, maybe those go into a faster release cycle.”
These ideas are circulating at a time when Kansas is preparing to release version 2.0 of its initial generative AI policy. The updated version will include more guardrails on how AI can be used by the state and will see more information collected on how business units are using the technology, he said.
As the state vets potential vendors’ security practices, it’s now starting to include questions about whether and how the companies use AI in their products and what will happen to government-owned data. It’ll be an ongoing process for the state to refine the approach and find the right balance of restrictions that can reduce risk without going too far and overly impeding “future business needs and capabilities.”
Godfrey is also paying attention to how AI can make disinformation attacks easier and more sophisticated. That includes by rapidly creating deepfakes.
As for how to stop deepfakes: “I don’t think you can,” Godrey said. As such, there needs to be focus on response and resilience.
