ChatGPT, Gemini, Copilot approved for use with Senate data

Staff in the upper chamber of Congress now have the go-ahead to use Senate data with three popular generative AI chatbots thanks to approval from an office that oversees the legislative body’s administrative operations. 

A recent notice from the Senate Sergeant at Arms’ chief information officer announced the approvals for Microsoft’s Copilot, Google’s Gemini, and OpenAI’s ChatGPT, expanding on previous policies. That memo was previously reported by the New York Times and independently obtained by FedScoop.

According to the document, Copilot is integrated into the Senate’s Microsoft 365 environment already, and more information about licenses for Gemini Chat and ChatGPT Enterprise will be coming within the next 30 days. Each Senate employee will be able to get one license for either Gemini or ChatGPT at no cost.

Approval of the tools comes as entities across the federal government — including Congress, executive agencies, and the federal judiciary — have been navigating their own use of the growing technology to reduce administrative toil and assist staff. 

The Senate, for its part, previously allowed ChatGPT, Google Bard, and Microsoft’s Bing AI chat in 2023 at “moderate” risk levels, but they were only for research and evaluation or use with non-sensitive data. The new approvals are less restrictive on the type of data that can be ingested, opening the door to more widespread use.

On Copilot specifically, the notice suggested it could be used to aid “routine Senate work, including drafting and editing documents, summarizing information, preparing talking points and briefing material, and conducting research and analysis.” 

The Microsoft chatbot can be used on mobile devices via the Copilot app or within environments like Word and Excel, the notice said. It also provided a link to a Copilot training for staff to learn more about the tool.

While much of the governance in executive agencies is shared publicly, the policies in Congress — as well as the judiciary — are more obscure. The Senate notice, for example, directs users to reference both office-level policies as well as the Senate’s AI policy, but that document isn’t public and it’s not clear how each office treats such tools.

In 2024, FedScoop found that some offices were experimenting with generative AI tools, but adoption wasn’t widespread. Individual office policies also varied in terms of maturity. But like everything in the rapidly evolving AI space, things could very well look different now. 

Aubrey Wilson, director of government innovation and global initiatives for the nonprofit PopVox Foundation — which has been chronicling the AI use in Congress for years — told FedScoop the tools are the first to be authorized for use with official Senate data under a governance framework established last year. 

According to Wilson, that policy created a two-tier system for risk assessment in October 2025: Tier 1 for non-sensitive Senate data and Tier 2 for official Senate data. The new approvals are the first under the latter tier, Wilson said. 

Like other policies in both the House and Senate, that framework was not made public. 

PopVox Foundation says its online tracking is based on its own review of House and Senate documents, but the lack of transparency in both chambers is something the organization, whose mission is helping legislatures use tech, wants to see change.  

“There’s no public list of the tools that are approved or what it takes to be approved in these chambers,” Wilson said. “And so because of that lack of transparency, it’s really difficult to have there be accountability or the ability to compare where the chambers are with their executive branch counterparts.”

The Senate Sergeant at Arms didn’t respond to FedScoop’s emailed requests for more information on the notice, description of the approval process for the tools, and details on how the tools are being procured. 

Notably, multiple AI developers are currently offering their services to federal agencies for deeply discounted rates of $1 or less for a year of access via the General Services Administration’s OneGov initiative. It’s unclear from the notice if similar deals have been extended to Congress. 

Spokespeople for OpenAI and Google didn’t provide responses to FedScoop’s emailed inquiries about the approvals or nature of the usage agreements. Microsoft declined to comment through a spokesperson. 

Although the details of the Senate policies are fuzzy, the notice did provide some insight into data availability through Copilot. The document says the chatbot doesn’t have access to Senate data unless the “information is explicitly shared within a prompt.” 

“Copilot does not search internal drives, shared folders, email, Teams chats, or any other Senate resources on its own,” the SAA notice said. “Copilot Chat operates in Microsoft’s secure government cloud and meets federal and Senate cybersecurity requirements.”