Microsoft said that the Department of Defense (DOJ) is leveraging the GPT-4 large language model (LLM) within an isolated and air-gapped Azure Government Top Secret cloud.
William Chappell, the Chief Technology Officer for Strategic Missions and Technologies at Microsoft reportedly conveyed that while AI models are connected to the internet, they are continuously compelled to defend against intrusive attempts and potential disruptions.
But when the chatbot is not connected to the internet it could be really useful for security agencies. It operates in an "air-gapped" environment, ensuring security and limiting access solely to the US government.
According to Chappell, Microsoft dedicated 18 months to developing the model. It is currently operational and is capable of responding to inquiries and even writing code.
While it can read and analyze files, it lacks the ability to learn from them to prevent sensitive information from entering the platform. He noted that the model is awaiting testing and accreditation by intelligence agencies.
Earlier this year, an official revealed that the Pentagon’s IT Division is developing an AI-powered chatbot like ChatGPT.
The Defense Information Systems Agency (DISA) is developing a technology that can take data from multiple sources, plug it into a database, and run a large language model on top of it.
The chatbot, however, is not designed to help defense personnel in mission-critical scenarios but to help assist captains with post-mission reports.
The Pentagon’s IT Division is in the process of developing a prototype that could be launched internally this year, according to reports.