SIAgentNews.com
← All news

Army to partner with OpenAI to automate software security vetting

Project Gatekeeper aims to use AI to speed up the process of approving commercial software for use on military networks.

· First reported by DefenseScoop

Papercut shadow box: picture for the story 'Army to partner with OpenAI to automate software security vetting'
AI-made papercut shadow box for this story. No real people or products are shown.
next quarterTimeline for integrating AI models

The short version

  • Project Gatekeeper aims to automate the 'authority to operate' (ATO) process for commercial software vendors.
  • The Army will use OpenAI's tools to help validate code submitted to a DevSecOps platform.
  • Vendors can currently use the system with a Defense Department Common Access Card, with plans to expand access.
  • The Army plans to eventually allow other open source and commercial model providers to join the initiative.

Why it matters

This move could significantly lower the barrier for private companies to sell software to the military. By automating security checks, the Army hopes to reduce months of bureaucratic delay while maintaining network security.

Streamlining military software

The U.S. Army is moving to modernize how it approves commercial software for use on its networks. Currently, the process of obtaining an authority to operate (ATO) is described as a burdensome task that can stretch on for many months. This delay often creates a significant barrier for private companies trying to sell their products to the government.

Project Gatekeeper was unveiled last month to address these specific bottlenecks. The initiative seeks to create a faster path for commercial vendors to get their software vetted and approved. By automating these checks, the Army hopes to create a more seamless transition from the commercial internet to secure military environments.

Acting CIO of the Army Gabe Chiulli explained that the goal is to provide a first initial vetting process. This would give vendors a "green check mark" indicating their code is good enough for military networks. This approach is designed to alleviate the relationship barriers that often exist between the government and industry partners.

How Project Gatekeeper works

The system works by allowing commercial vendors to submit their code directly to an Army DevSecOps platform. This platform is designed for quick validation of software components. Once the code passes the necessary checks, it can be moved into an IL5 environment, which is a secure enclave for distributing products.

To facilitate this, the Army is integrating AI models into the validation pipeline. The first phase of this integration involves a partnership with OpenAI to use tools like Daybreak. These tools will help the Army perform automated checks on the submitted code to ensure it meets security standards.

The Army is being intentional about how it uses these AI models to avoid redundant work. Chiulli noted that they do not want to use tokens just for the sake of using them. Instead, they want to apply AI in a strategic way to check for specific vulnerabilities and security requirements that are already known to the military.

Access for commercial vendors

Currently, vendors who possess a Defense Department Common Access Card can log in to the Gatekeeper system to begin the process. However, the Army is working on expanding this access to a broader audience. Officials are looking into allowing companies to access the system via the Login.gov platform to make it more accessible to the private sector.

The Army also plans to establish reciprocity with other military departments. This means they want to create a system where different branches of the military can share the same accreditation results. By doing this, the military can avoid building separate, redundant systems for the same type of software validation process.

For companies that want to maintain their own privacy, the Army plans to encourage them to bring their own API keys. This setup ensures that the companies remain the only ones who can see the logs on their code. This approach allows vendors to pay for their own security while still participating in the Army's automated vetting framework.

The roadmap for AI integration

The Army is taking a phased approach to introducing AI into this software vetting pipeline. The initial phase is focused on working with OpenAI to determine which specific checks the military wants to prioritize first. This allows the Army to refine the process before opening it up to a wider variety of models.

Following the initial work with OpenAI, the Army intends to open the project to other commercial and open source model providers. They will be looking for different ways and harnesses to check various types of software requirements. This expansion will allow for a more diverse range of tools to be used in the validation process.

The ultimate goal is to create a system that is both secure and efficient for the commercial internet. By using AI to handle the heavy lifting of code scanning, the Army hopes to provide a clear path for innovation. This transition aims to move away from manual, months-long reviews toward a more automated and scalable security model.

What to watch

  • How quickly the Army can successfully move code into IL5 environments.
  • Which specific security checks the Army chooses to prioritize in the first phase with OpenAI.
  • Whether other military branches successfully adopt the reciprocity model.

Try it

Look up the Login.gov platform to see how the Army plans to eventually host the gateway for commercial vendors.

Sources and links

Read the original at DefenseScoop

First reported by DefenseScoop. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.

Liked it? Say why

Add a line for other readers.

No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.

More AI news