U.S. Department of Defense Accelerates AI Deployment, Google Gemini Agent to Enter Office Systems

The Deputy Secretary of Defense for Research and Engineering, Emil Michael, revealed that Google is planning to introduce AI agents to 3 million Department of Defense employees to automate daily tasks.

Michael stated that Google’s Gemini AI agents can independently complete tasks after users set them up. Initially, they will operate on “non-classified” networks. “We are starting with non-classified networks because most users are non-classified, and then we will expand to classified and top-secret levels.”

Michael added that the Pentagon is discussing the possibility of deploying these AI agents on classified cloud networks. “I am very confident they will become excellent partners across all our network platforms.”

On Tuesday, March 12, Jim Kelly, Vice President of Federal Sales at Google Public Sector, announced on the official blog a new feature for Gemini for Government on the GenAI.mil platform—Agent Designer.

Kelly wrote, “Now, with Agent Designer, Department of Defense civilian and military personnel can build their own agents to support non-classified tasks.” “Users can create digital assistants without coding, automating repetitive and multi-step administrative tasks.”

Source: Google Official Blog

Recently, the U.S. government has expanded its use of AI, sparking controversy within many American tech companies developing this technology. After the U.S. and Israel began military strikes on Iran, news of the U.S. using the Claude model during this operation was disclosed.

In response, Anthropic, the developer of Claude, insisted that it refused to lift two constraints requested by the Department of Defense: preventing the model from being used for large-scale domestic surveillance and for fully autonomous weapons systems without human intervention. This stance has led to a standoff between Anthropic and the U.S. Department of Defense.

Last week, the Department of Defense officially designated Anthropic as a “supply chain risk” (barring it from obtaining any government contracts). Michael also accused Dario Amodei, CEO of Anthropic, of wanting to play God.

It should be noted that Google also faced internal protests over its collaboration with the Department of Defense. In 2018, thousands of employees opposed the company’s involvement in “Project Maven,” which aimed to use AI to analyze video data from overseas drone warfare. Google ultimately did not renew the contract for that project.

However, Google later relaxed some restrictions on military collaborations. Michael stated that Google is now a “trusted and supportive partner.”

According to Michael, Google will launch eight ready-made AI agents in the latest collaboration, mainly for automating administrative tasks such as summarizing meeting notes, preparing budgets, and checking action plans. Some agents may also have operational impacts, helping to plan missions and estimate resource needs.

Michael emphasized that the Pentagon should expand AI applications. “When I took over the AI project here in August, I was somewhat shocked because we didn’t have the basic AI capabilities that most people and consumers around the world now have.”

He said that training AI agents by Department of Defense personnel is crucial to prevent errors, but the current training progress is far behind actual deployment. “It can save you a lot of time, but ultimately, you need to review it to ensure there are no hallucinations.”

(Source: Caixin)