Microsoft's new AI model can run 'secretly' on government-operated networks: Key details

According to a new Bloomberg report, Microsoft has deployed a new generative AI model for US intelligence agencies to analyse top-secret information. This AI model is based on ChatGPT-4 and would be disconnected from the internet. It is the first time Microsoft has deployed a major language to allow spy agencies to decode secret information.

William Chappell, Microsoft’s strategic missions and technology CTO, that Microsoft spent 18 months working on this model, and it’s currently air-gapped and accessible to the US government only. Further, it will be also able to answer questions and write different codes.

He also clarified that it can read and analyse files but cannot learn from them to stop sensitive information from entering the platform. Therefore, this AI model is yet to be tested or authorised by the intelligence agencies. While this AI model is technically cloud-based, it's a static cloud that's isolated from the internet, so it can't learn from data and files shared by users from the web.

This new AI model was activated on Thursday and is currently available to about 10,000 individuals in the intelligence community, with plans to increase its user base. But, it is certainly the first ever AI model built specifically for classified workloads, as well as the first major large language model to operate fully separate from the internet.