Pentagon Officials Say AI Accelerates Cyber Defense and Exploitation

Top Pentagon technology officials voiced optimism Thursday about the potential for AI models to strengthen cyber defenses by finding and patching vulnerabilities far faster than humans can manage. Katherine Sutton, Assistant Secretary for Cyber Policy, spoke at the SCSP AI+Expo this morning. She said she hears a lot of people talking about challenges and threats when they talk about Mythos.

"[But] there’s huge opportunity in these models. One of the foundational things that they’re going to enable is the development of secure code," Sutton stated. " Emil Michael, the Pentagon’s CTO and Under Secretary for Research and Engineering, spoke at the SCSP AI+Expo immediately after Sutton.

David Sanger moderated the session. "Those vulnerabilities have always existed. Now you can find them faster, and the good news is you can patch them faster.

So that creates a moment where it’s incumbent upon the country, not just the government, to harden the systems," Michael said. Michael described how the Defense Department and the nation rely on a patchwork of often-antiquated software systems running sloppy, buggy code.

He argued that AI capable of patching code autonomously could dig America out of decades of accumulated tech debt and fast-forward what should have been done for the last 20 years.

The officials' remarks came against the backdrop of an ongoing dispute with Anthropic. Anthropic is pursuing two parallel lawsuits against the government. President Donald Trump and Secretary Pete Hegseth banned federal use of Anthropic products in blistering social media posts.

Michael himself called Anthropic’s CEO "a liar [with] a God-complex" on X. ” He described Mythos as really just an example of the upcoming evolution of cyber-capable models, just slightly ahead. "These kinds of cyber models … whether OpenAI’s or xAI’s, Google’s … will all come out in the next year or so [with] this exquisite cyber capability.

Hopefully, they have mostly the same data, and they’re mostly going to converge," he said. Sutton agreed the focus extended beyond any single system. "This isn’t about this one single model.

However good Mythos may be today, in six months, there’s going to be something that is even better," she said. "These models are coming one way or the other. Thankfully, it’s the US companies that are that are ahead, not the Chinese ones," Michael added.

OpenAI made its rival to Anthropic’s Mythos more widely available to cyber defenders last week. Eight leading tech firms received clearance last week to deploy their AI on classified networks. " Top Pentagon tech officials are optimistic that Mythos-style AI tools will improve cyber defense.

An image released by the Defense Department showed a 149th Intelligence Squadron airman conducting training in a computer training lab at Mather Field, California on December 2, 2023.