Elon Musk's team at the US government used Meta's Llama 2 AI model to analyze civil servants' responses to a "voluntary resignation" email, Wired reports. Despite rumors, Musk did not use his own Grok AI model.
In January, the newly created Department of Government Efficiency (DOGE), led by Elon Musk, sent a letter to civil servants titled "Fork in the Road," suggesting they resign by simply replying "resign." Musk had already used a similar approach after buying Twitter in 2022. DOGE used a local Llama 2 startup to process the responses, without transmitting data to the internet.
DOGE operates within the Office of Personnel Management (OPM), which serves as the HR service for all federal agencies. Among the department's first initiatives is the launch of government mail and the automation of redundancies. The technical infrastructure for the mailings was developed by former Tesla engineer Riccardo Biasini.
In February, civil servants began receiving new letters demanding that they submit five-point reports every week. This raised concerns about data security, as officials suspected that AI was processing the responses. While Wired was unable to confirm the use of Llama 2 for the analysis, many employees believe that it was the one processing the reports.
Although Elon Musk’s xAI company developed its own Grok model, DOGE did not use it to launch the postal system, as the model was only available in closed mode at the time. Microsoft recently added support for Grok 3 on the Azure platform, opening up opportunities for its use in the public sector. Earlier this year, Palantir also integrated Grok into its solutions, which are already used by government agencies.
DOGE has also been linked to other AI tools in government: AutoRIF for automated staff reduction, the GSAi chatbot based on Meta and Anthropic models, and CamoGPT, which the Army used to clean up DEI references in training materials.