AI Model Examines Federal Workers' Emails Utilizing Doge Technology
Elon Musk's Department of Government Efficiency (DOGE) utilized Meta's Llama 2 AI model to analyze emails from federal workers, as revealed by materials obtained by our news site. Affiliates of the DOGE, stationed within the Office of Personnel Management (OPM), tested and implemented Meta's Llama 2 model to review and classify responses to the controversial "Fork in the Road" email that was circulated in late January across the government.
The email, reminiscent of one sent by Musk to Twitter employees after he took over the company in 2022, offered deferred resignation to those opposed to changes being made to the federal workforce, including an enforced return-to-office policy, downsizing, and a requirement to be "loyal." To resign, recipients merely had to reply with the word "resign."
Evidence indicates that Meta's Llama model was deployed to sort through email responses from federal workers to determine the number of acceptances. The model is believed to have been run locally, reducing the likelihood of data transmission over the internet.
Meta and OPM did not respond to our requests for comment.
The connection between Meta and governmental tech usage became more apparent when Meta CEO Mark Zuckerberg, along with other Silicon Valley tech leaders like Musk and Amazon founder Jeff Bezos, attended Trump's inauguration in January. However, until now, little was known about Meta's technology being used in the government. Given Llama's open-source nature, the tool can conveniently be used by the government to support Musk's reform initiatives without explicit consent from the company.
After Trump took office in January, DOGE operatives reportedly infiltrated OPM, an independent agency functioning as the human resources department for the federal government. The new administration's main objective for the agency was to develop a government-wide email service, according to current and former OPM employees. Riccardo Biasini, a former Tesla engineer, is said to have been involved in constructing the infrastructure for the service, which eventually sent out the original "Fork in the Road" email.
In late February, weeks after the "Fork" email, OPM sent out another email to all government workers requesting them to submit five bullet points outlining their weekly accomplishments. These emails reportedly created a state of chaos within various agencies as employees grappled with securing sensitive information in their responses, some of which had turned on read receipts, allegedly discovering that their responses were not being opened. NBC News reported at the time that these emails were expected to be analyzed using an AI system, though our materials do not explicitly indicate DOGE affiliates using Meta's Llama models to analyze these weekly submissions. However, two federal workers inform our news site that it would not be difficult for them to do so.
- Elon Musk's Department of Government Efficiency (DOGE) is using Meta's Llama 2 AI model for political analysis, as seen in their email review process.
- The DOGE, within the Office of Personnel Management (OPM), employs Meta's Llama 2 model to classify responses to controversial emails, such as the "Fork in the Road" email.
- The use of artificial-intelligence, specifically Meta's Llama model, has been deployed to sort through email responses from federal workers, overseeing deferred resignations and policy changes.
- The connection between tech companies like Meta and governmental policy-and-legislation is becoming evident, with newsletters highlighting this relationship and its implications.
- The social impact of technology is under scrutiny, as the use of open-source AI tools like Llama can support policies and reforms without explicit company consent, raising questions about data security and privacy.
- Weeks after the "Fork in the Road" email, the Office of Personnel Management (OPM) sent out another email requesting weekly accomplishments from government employees. It remains unclear if DOGE affiliates utilized Meta's Llama models for AI analysis of these submissions, but two federal workers indicate it would be possible.
