Advertisement

Army plans to work with OpenAI on Project Gatekeeper

Acting CIO Gabe Chiulli provided an update Friday on the Army’s cybersecurity efforts.
Listen to this article
0:00
Learn more. This feature uses an automated voice, which may result in occasional errors in pronunciation, tone, or sentiment.
Photo illustration of U.S. Army cyber soldier in operations center. (U.S. Army photo by Tài Doick)

The Army plans to work with OpenAI as it moves forward with a new initiative to accelerate the accreditation process for commercial vendors looking to offer their software to the military, according to the service’s chief information officer.

Project Gatekeeper, which was unveiled last month, aims to automate the traditionally burdensome authority to operate (ATO) process, which often stretches on for months, allowing commercial vendors instead to submit code to an Army DevSecOps platform for quick validation.

“What Gatekeeper was all about was taking that same experience that we built on the DSO platform, bring it to a commercialized space and say, hey, we’ll get you that first initial vetting that says you get the green check mark, that you are actively good enough for our networks. Get their code through, and then that kind of alleviates that relationship barrier between industry that doesn’t have a relationship with the government,” Acting CIO of the Army Gabe Chiulli said Friday at the Annual Data Forum, presented by AFCEA International.

“They get the government check, and that code gets moved into an IL5 environment after it gets through all the different checks, and says, ‘hey, you are now capable of essentially selling your product or distributing your product into a secure Army enclave,’” Chiulli added.

Advertisement

He said he’s working with the other military department CIOs to try to establish “reciprocity” so that they don’t have to build their own systems for this type of process.

Vendors with a Defense Department Common Access Card can log in to Gatekeeper to use the system. But, Chiulli noted, officials are working to open it up more widely and allow companies to access it via the Login.gov platform.

At Friday’s conference, DefenseScoop asked the Army CIO about the timeline for integrating AI models into the Gatekeeper initiative.

“I’m going to put it into the next quarter,” Chiulli replied.

“I think what it’ll be is a phased approach of different commercialized models. So remember, this is on the commercial internet. The first phase of this was always meant to be that we were going to work with OpenAI in the first phase because … we’re using something like Login.gov. It helps us do the validation” using tools like OpenAI’s Daybreak, he said.

Advertisement

The Army wants to be cautious about how it releases an AI model vetting process into the DevSecOps pipeline, he noted.

“So, the first phase is working with OpenAI to kind of work through, like, what do we want to check first? Because there’s a lot of good tooling out there that does validation checks on software already, so we don’t want to repeat and just use tokens just for tokens purposes. We want to apply it in the most strategic, smart, intelligent way,” Chiulli said.

OpenAI did not immediately respond to a request for comment Friday.

After the first phase, the Army will likely start opening it up to other open source and commercial model providers, Chiulli said.

He noted that he’ll encourage companies to bring their own API keys.

Advertisement

“When you think about code scanning in an API era in a vulnerability scanner, it’s going to have logs on their code. So we want to make sure that they’re the only ones that can see that … from that perspective. So that is a protective kind of way we’re thinking through this so that, you know, as they bring their own API key they’re paying for that security from their perspective,” Chiulli said.

“We’ll do a call out at some point to … request different ways and different harnesses, if you will, for checking different things,” he added.

Latest Podcasts