(March 11): Alphabet Inc’s Google is introducing artificial intelligence (AI) agents across the Pentagon’s three million-strong workforce to automate routine jobs, according to a senior defence official.
Google’s Gemini AI agents, which can undertake work independently on behalf of a user who sets them tasks, will initially operate on unclassified networks, said Emil Michael, the under secretary of defence for research and engineering.
“We are starting with unclassified because that’s where most of the users are, and then we will get to classified and top secret,” Michael said in an interview, adding that talks with Google over using the agents on the classified cloud are underway. “I have high confidence they are going to be a great partner on all networks.”
The new feature will allow civilian and military personnel at the Defense Department to build AI agents using natural language, Google vice-president Jim Kelly said in a blog post on Tuesday. A Google Cloud spokesperson didn’t respond to a request for comment about whether it is negotiating to put AI agents on the classified cloud.
The military’s expanding use of AI is fuelling controversy at many of the cutting-edge American companies developing the technology. In the Iran war, the US has used AI to help identify targets and speed processes, allowing for the unprecedented intensity of the bombing campaign.
The Pentagon has been bringing online more products from the likes of OpenAI and Google as part of a push to speed AI adoption in the military. That effort triggered its escalating feud with Anthropic PBC, which sought guardrails on using its technology for domestic surveillance and in fully autonomous weapons. In response, the Defense Department labeled the company a supply-chain risk last week.
See also: Nintendo shares gain US$14 bil with smash hit ‘Pokopia’
Anthropic is suing the government over the designation, which is typically reserved for companies from countries the US views as adversaries.
Michael, who led negotiations with Anthropic, said the issue wouldn’t be resolved through the courts and that the Pentagon is now “moving on”.
Until recently, Anthropic provided the only AI system that could operate in the Pentagon’s classified cloud. The Defense Department has in recent weeks struck deals with OpenAI and xAI to operate on restricted networks, and Michael touted its expanding AI cooperation with Google.
See also: Singapore must train more people to build AI, top official says
Google has in the past faced internal criticism over its ties with the Defense Department. In 2018, thousands of Google employees protested the company’s involvement in Project Maven, a Pentagon effort to develop AI to analyse video feeds taken in America’s overseas drone wars. The backlash led Google to not renew its contract for the programme.
Google later dropped some prohibitions on working with the military. Michael said the company is a “trusted” and “supportive” partner.
Initially, Google will roll out eight ready-made agents, some of which will focus on automating jobs such as summarising meeting notes, creating budgets and checking proposed actions against the tenets of the national defence strategy, according to Michael. Some AI agents on the unclassified network could have operational impact, helping with planning and resourcing estimates for military tasks and operations, according to briefing notes provided by a Pentagon spokesperson.
Users can also design their own digital assistants to automate repetitive, multi-step administrative tasks without needing to know any code, according to the Google blog post.
Michael said training, guidance and policies could help alleviate the potential risks associated with introducing AI into the workforce, such as when agents magnify or mask errors.
“It saves you a lot of time in the middle, but you have to review at the end to make sure there’s no hallucinations,” he said, adding that the Pentagon should expand AI usage. “When I got here and took over the AI portfolio in August I was somewhat shocked that we didn’t have the basic AI capabilities that most people, consumers around the world have now.”
Google’s AI chatbot accessible through the GenAI.mil portal has been used by 1.2 million Defense Department employees for unclassified work since it was launched in December. It will offer Gemini agents from Tuesday. Defence personnel have run 40 million unique prompts and uploaded more than four million documents, according to a Pentagon spokesperson.
To stay ahead of Singapore and the region’s corporate and economic trends, click here for Latest Section
However, training — which Michael said was critical for preventing errors — far lags usage. Only 26,000 people have been trained in how to use AI since December and future sessions run by the Defense Department are fully booked, according to the spokesperson.
The AI portal is already streamlining planning for large-scale military simulations, according to Kenneth Harvey, the director for the Mission Training Complex at Fort Bragg and the Army’s rapid-reaction 18th Airborne Corps.
Harvey, who has been experimenting with the AI platform for a couple of months, said the process of designing military exercises for as many as 50,000 simulated soldiers in which US forces would help defend a Baltic country took his nine-person staff six months.
It took six weeks to plan a similar exercise scenario focused on US Southern Command using the AI portal, Harvey said, adding “human eyes vetted every word”.
Uploaded by Tham Yek Lee
