Musk's "Department of Government Efficiency" (DOGE) uses Meta Platforms open source AI models to assist in office work, raising concerns about data security and privacy

Mars Finance News, on May 23, a material reviewed by the famous online magazine Wired shows that Doge, belonging to the Office of Personnel Management (OPM), tested and used Meta's Llama 2 model to classify and organize the content of an email titled "Fork in the Road" for federal employees. It is worth noting that although Meta has launched more advanced Llama 3 and Llama 4 models, Doge still used Llama 2. Elon Musk's AI model Grok was not yet open in January, so it could not be used by Doge to handle the task at that time, as Grok was still a proprietary model and not yet available for external use. Wired reported that Doge may rely more frequently on the Grok model in the future. The Verge pointed out that Microsoft announced this week that it will host xAI's Grok 3 model on its Azure AI Foundry platform, which means these models will be open to the public and able to support more practical uses. Legislators emphasized that Doge's use of AI systems to analyze federal employees' emails raises "significant security concerns" and lacks transparency. More than 40 legislators wrote to the Director of the Office of Management and Budget, Russell Vought, in April, requesting an investigation into Doge's AI usage. They warned that this approach could pose "serious security risks" and potentially "hinder the appropriate adoption of AI". US legislators are also concerned that Musk might increasingly use his own models, potentially benefiting from government data inaccessible to his competitors, while also risking data leakage. (Jinshi)