DOGE tested Meta’s Llama 2 AI model to check responses from federal employees to the “Fork in the Road” email, raising concerns about its legality. The federal initiative under US President Donald Trump’s administration, the Department of Government Efficiency (DOGE), has recently used artificial intelligence to sort emails.
DOGE Used Llama 2 in January to send an email to federal employees, with the subject of “Fork in the Road” to discuss new workforce reforms. The email put forward the US government’s next policies and agenda. It explained downsizing and demanded loyalty to the government. Otherwise, they can reply with the word “resign” to opt into the deferred resignation program.
What is the deferred resignation program?
In January 2025, the US government offered to all federal government employees to leave their positions. The government will continue paying their salaries and benefits until September 30, 2025. In this program, voluntary resignations are encouraged to avoid the harms of federal workforce restructuring.
What is Meta’s Llama 2 AI Model?
Meta’s Llama 2 AI model is a Large Language Model (LLM) that helps to perform various tasks, including email classification. It is an open-source LLM that supports research and commercial use.
How DOGE Use Llama 2 to filter emails?
To determine the number of employees who have resigned, the US government has integrated the Meta Llama 2 AI model. The AI model can collect replies and understand tone, patterns, and intent. As per its training, it evaluates and categorizes specific responses.
Elon Musk’s role in the Trump administration
Since Donald Trump took office, Elon Musk, CEO of TESLA helped the US government in DOGE development. Later it joined the Office of Personnel Management (OPM) to manage federal employees. Musk did not officially join the government job but helped the government with revolutionary tech ideas. On the other hand, confirmed sources have told last night that Elon Musk has left the Trump administration due to uneven performance.
Why did DOGE not use Grok AI?
In the development of the government email system, DOGE did not use Grok AI in Elon Musk’s DOGE project because earlier Grok AI was not available for free. Moreover, it had limited features at that time and there were only limited methods to test or connect to it.
AI model analyzing emails
DOGE Used Llama 2 to categorize the responses of federal employees who accepted the deferred resignation program. The AI model was running locally which means it was not available on the internet and only ran on government computers.
Is it legal to use AI to review resignations?
According to the United States Federal guidelines, reviewing resignations is legal but it must comply with employment laws. For example, AI must not discriminate based on race, age, or gender.
US employees raised privacy concerns.
In February, US employees were asked to submit five bullet points about their weekly progress. Such instructions from the US government created chaos among the government employees. Because their responses contained sensitive information. According to a report, many employees who use read receipts said that no one opened their responses.
According to NBC News, it is expected that the information within emails will go to the AI system. However, there is no verified information that DOGE used Llama 2 to analyze these five bullet points. Like the AI model, it was used to categorize Fork emails.
