Microsoft Corp. will make it possible for users of its Azure Government cloud computing service, which include a variety of US agencies, to access artificial intelligence models from ChatGPT creator OpenAI.

Microsoft, which is the largest investor in OpenAI and uses its technology to power its Bing chatbot, plans to announce Wednesday that Azure Government customers can now use two of OpenAI’s large language models: The startup’s latest and most powerful model, GPT-4, and an earlier one, GPT-3, via Microsoft’s Azure OpenAI service. The Redmond, Washington-based company plans Wednesday to release a blog post, viewed by Bloomberg, about the program, although its doesn’t name specific US agencies expected to use the large language models at launch. The Defense Department, the Energy Department and NASA are among the federal government customers of Azure Government.

The Defense Technical Information Center — a part of the Defense Department that focuses on gathering and sharing military research — will be experimenting with the OpenAI models through Microsoft’s new offering, a DTIC official confirmed.

Microsoft already offers OpenAI models to its commercial clients, with the Azure OpenAI service growing rapidly in recent months. Microsoft said in May it had 4,500 customers for the service, a jump from 2,500 the previous quarter, including Volvo AB, Ikea, Mercedes-Benz Group AG and Shell Plc. The initiative announced Wednesday is the first known effort by a major company to make the chatbot technology widely available to the US government.

Interest in large language models, which are trained on huge swaths of internet data so that they can predict and generate humanlike responses to user prompts, has skyrocketed since the public release of OpenAI’s ChatGPT chatbot in late 2022. Since then, tech companies large and small have begun offering powerful chatbots to users and conversations have erupted in Congress regarding whether and how artificial intelligence should be regulated.

Federal, state and local government customers can access OpenAI’s GPT-4 and GPT-3 models for tasks such as generating answers to research questions, producing computer code and summarizing field reports, Microsoft’s Bill Chappell wrote in the blog post. Though they will be able to use the models with a chat-like interface, Azure Government users won’t have access to ChatGPT specifically, a Microsoft spokesperson confirmed. ChatGPT is generally available via the Azure OpenAI service. Chappell previously worked for the US Defense Advance Research Projects Agency, or DARPA.

Microsoft hosts the OpenAI models in its commercial cloud computing space, which is separate from the cloud used by Azure Government customers that adheres to a variety of specific security and data-compliance rules.

Chappell wrote that because Microsoft hosts the models in its Azure infrastructure, any data sent to them stays within the Azure OpenAI service, and added that data from Azure Government customers won’t be used to train the AI models.