Microsoft on Thursday launched a generative AI model for US intelligence agencies to analyze top secret information, Bloomberg reports. Microsoft announced the AI offering on Tuesday at the 2024 SCSP AI Expo for National Competitiveness in Washington, DC. The GPT-4 based model is designed to be completely separated from the Internet and to securely handle sensitive data.
“This is the first time we have an isolated version – isolated meaning it is not connected to the Internet – and it is on a special network that only the US government can access,” said William Chappell, Chief technology officer for strategy at Microsoft Missions and Technology, said Bloomberg. “You don’t want it to learn from the questions you ask and then somehow reveal that information.”
Intelligence communities have desperately sought a ChatGPT-like product, but the security risks with generative AI products are too high. Chappell tells Bloomberg that this model, introduced last week, is secure thanks to an “air-gapped” cloud environment that isolates it from the internet. While most AI models learn from the files uploaded to them, this GPT-4 spy model does not. This allows the government to keep this model “clean” and prevent classified information from leaking into the platform.
This appears to be the first AI model designed specifically for classified workloads, as well as the first large-scale language model to operate completely independently of the Internet. Last year, the Central Intelligence Agency has introduced its own ChatGPT-style AI tool to search through reams of public information. However, this AI model has not been used in any classified documentation. Other government agencies, including the Pennsylvania Office of Administrationuse generative AI to work with unclassified data.
The CIA and the Office of the Director of National Intelligence did not immediately respond to Gizmodo’s request for comment.
Microsoft has reportedly been working on this GPT-4 espionage model for the past 18 months, which included overhauling an existing AI supercomputer in Iowa. According to Bloomberg, the model has been in operation for less than a week and has yet to be tested and accredited by the intelligence community.
This introduction of an AI model for top-secret workloads comes at a critical time for Microsoft’s cybersecurity reputation. The Department of Homeland Security recently issued one scathing review of Microsoft’s security practices, accusing the cloud provider of exposing the emails of high-ranking government officials. The report states that Microsoft’s security culture “needs an overhaul.” This prompted CEO Satya Nadella to recently tell employees that security was now Microsoft’s “business.”highest priority.”
At the same time, Microsoft is not afraid to develop AI tools for government agencies. An April report from The Intercept revealed that the Microsoft Azure version of DALL-E was this used as combat equipment for the US Department of Defense. Microsoft has a long history of working with militaries and government agencies, although its adoption of AI is relatively new.