GSA’s plans to test the controversial AI tool Grok; Why IRS’s data-sharing deal with ICE could lead to ‘dangerous’ mistakes

Employees at the General Services Administration appear poised to test Grok 3, the artificial intelligence tool built by Elon Musk’s company xAI, according to a GitHub page referencing the agency’s work. The GitHub page operated by GSA and its digital government group Technology Transformation Services references the Grok AI model as one it is testing and that the team is actively discussing as part of its 10x AI Sandbox. A GSA spokesperson told FedScoop in a response to an inquiry about the agency’s work with Grok “GSA is evaluating the use of several top-tier AI solutions to empower agencies and our public servants to best achieve their goals. We welcome all American companies and models who abide by our terms and conditions.”A post from Tuesday shows what appears to be one GSA employee trying to access Grok 3 for testing, but struggling to do so. Several names of the people active on the GitHub page match those of workers affiliated with GSA. The 10x AI Sandbox project is described on GitHub as “a venture studio in collaboration with the General Services Administration (GSA). Its primary goal is to enable federal agencies to experiment with artificial intelligence (AI) in a secure, FedRAMP-compliant environment.” It continues: “By providing access to base models from leading AI companies and offering advanced UI features, the sandbox empowers agencies to test and validate new AI use cases efficiently.” The public version of the 10x AI Sandbox project page on GitHub was taken down after the publication of this story, redirecting now to a 404 error page. Interest in testing Grok comes as GSA continues to work on GSAi, an artificial intelligence tool built by the agency and meant to help employees access multiple AI models. At launch, the GSAi tool included access to several systems, including tools from Anthropic and Meta. Notably, Grok came under fire last week after promoting various antisemitic statements on the Musk-owned social media platform X.

A top digital rights group is pushing back on the IRS’s data-sharing agreement with the Department of Homeland Security, writing in a new court filing that the pact violates federal tax code and fails to take into account the real-world consequences of bulk data disclosure. In an amicus brief filed in the U.S. Court of Appeals for the D.C. Circuit, the Electronic Frontier Foundation argued that the “historical context” of the tax code section that ensures confidentiality of returns and return information “favors a narrow interpretation of disclosure provisions.” EFF also made the case for why the bulk disclosure of taxpayer information — in this case to Immigration and Customs Enforcement — is especially harmful due to “record linkage errors” that set the stage for “an increase in mistaken and dangerous ICE enforcement actions against taxpayers.” Nonprofit groups sued the Trump administration in March, shortly after the data-sharing deal between the IRS and ICE was announced. Soon after, the tax agency’s then-acting commissioner resigned, reportedly in protest. In May, a Trump-appointed federal judge refused to block the agreement, allowing the IRS to continue delivering taxpayer data to ICE. The ruling, DHS said in a statement, was “a victory for the American people and for common sense.” As the D.C. Circuit Court considers the appeal, the Electronic Frontier Foundation wants to make sure that the “historical context” of tax and privacy law is taken into account.

The Daily Scoop Podcast is available every Monday-Friday afternoon.

If you want to hear more of the latest from Washington, subscribe to The Daily Scoop Podcast  on Apple Podcasts, Soundcloud, Spotify and YouTube.