Microsoft Develops Air-Gapped AI for U.S. Intelligence Agencies

Microsoft has developed a new artificial intelligence (AI) model that is not directly connected to the internet. This “isolated” AI service will allow U.S. intelligence agencies to securely use the power of AI to analyze top-secret information.

A Microsoft executive said this is the first time a large language model has operated entirely independently of the internet. Most AI models, such as OpenAI’s GPT, rely on the internet and cloud servers to learn new information and extract patterns from data. However, Microsoft wanted to build a “truly secure” system for the U.S. intelligence community.

“Spy agencies around the world want AI to help them understand and analyze the classified information they produce on a daily basis,” Bloomberg writes. Internet-connected AI could leak sensitive information, so intelligence organizations have limited use of this technology.

Microsoft reportedly developed its new AI model over the past 18 months based on GPT-4 and used a supercomputer to build it. Bloomberg, citing a Microsoft electrical engineer, writes: “The new AI is on a special network that only the US government has access to.”

Microsoft’s new AI is not “dynamic,” meaning it can review data but does not learn from it. This way, the U.S. government can ensure that sensitive information remains confidential. According to the Microsoft electrical engineer, theoretically 10,000 people have the ability to access the AI in question.

Back to top button