eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
With the ongoing AI arms race between the US and China, it’s not far-fetched for the Pentagon to use generative AI in military operations.
The Department of Defense is moving into the second phase of military AI adoption, leveraging generative models to support ground operations and intelligence analysis. This effort builds on earlier AI breakthroughs to assist field personnel. The first phase began in 2017 with Project Maven, which used computer vision to analyze field photos and satellite images.
Using generative AI in the military
When the military uses generative AI, it can gather and interpret intelligence, electronic signals, and reconnaissance information, helping soldiers gain situational awareness.
The military can leverage AI for task-specific or mission-oriented efforts, enhancing productivity and efficiency. It can make decisions based on the data it analyzes, including specialized information, military tactics, strategic manuals, and real-time surveillance data.
AI can handle repetitive tasks like drafting emails, writing reports, and analyzing data, helping reduce the workload on military personnel. It can also generate operations orders and identify potential threats by analyzing surveillance inputs.
Possible downside of AI and the futility of ‘human in the loop’
However, the military’s use of generative AI has raised ethical concerns.
For instance, human rights groups have expressed concerns about allowing AI to generate target lists or suggest actions based on intelligence data. To mitigate these concerns, defense agencies often require human oversight in AI decisions — also called having a “human in the loop.” However, given the vast scale of data involved, this oversight may not be practical.
Heidy Khlaaf, safety expert specializing in autonomous software and chief AI scientist at the AI Now Institute, said, “‘Human in the loop’ is not always a meaningful mitigation.” She continued, “It wouldn’t really be possible for a human to sift through that amount of information to determine if the AI output was erroneous.”
AI models operate at a level of data complexity that humans can’t easily replicate. Generative AI, in particular, relies on data, which could replace legacy intelligence analysis approaches.
Classifying military data
Large language models (LLMs) excel at analyzing vast amounts of data, identifying patterns, and connecting seemingly unrelated details in military documents. LLMs can classify data with compelling accuracy, potentially solving the Pentagon’s long-standing issue of under-classification and over-classification of sensitive materials.
Chris Mouton, a senior engineer for RAND, said, “I don’t think anyone’s come up with great answers for what the appropriate classification of all these products should be.”
Pentagon partners with defense tech companies
The military has partnered with defense tech companies such as Anduril, Scale AI, and Palantir to use military hardware and AI to collect, manage, and interpret surveillance and intelligence data. Palantir collaborates with Microsoft to offer AI tools for military use. These AI systems are trained using classified military datasets.
For instance, Vannevar Labs, a defense tech company, provided the U.S. military with generative AI tools in November 2024, securing a $99 million contract with the Pentagon’s Defense Innovation Unit. Vannevar’s AI models can translate data from foreign sources, detect potential threats, and analyze political sentiments.
“Our real focus as a company,” said Scott Philips, Vannevar Labs’ chief technology officer, is to “collect data, make sense of that data, and help the US make good decisions.”
Military AI initiatives accelerated under the Biden administration, expanding the scope of early programs such as Project Maven. The tense U.S.-China race for technological dominance and increasingly complex geopolitics will further speed up AI adoption for military operations.