AI and autonomy in contested operations
Bottom lines up front
- The US military should presume that, from now on, most future operations will occur in contested environments where connectivity is poor or nonexistent.
- Acquiring military AI capabilities for such environments is one challenge; building servicemembers’ trust in and ability to use them responsibly is a related but distinct challenge.
- The private sector’s experience managing the rollout of autonomous capabilities and incorporating user feedback into subsequent updates offers valuable lessons.
The next major conflict the United States fights will be fought on a battlefield reshaped by artificial intelligence. Among the other adjustments this will require, US forces should assume their systems will operate in denied, degraded, intermittent, and limited (DDIL) conditions. This could include jamming, cyber attacks, and anti-satellite weapons, all aimed at cutting connectivity and throwing operations into chaos.
AI could help forces fight through that disruption, from autonomous systems that keep track of targets after losing contact with their operator to decision-support tools that help servicemembers piece together a fractured picture. However, the technology is outrunning the authorities, testing practices, and training needed for servicemembers to trust it at the tactical edge. That gap is exactly where US adversaries will strike. This report draws on the Pentagon’s drone-dominance and counter-UAS programs, iterative efforts like Thunderforge, simulation environments like the Marine Corps’ Project Tripoli, and private-sector lessons from John Deere and Tesla on earning user trust in autonomous systems, to lay out where policy and practice have fallen behind.
What will it take to catch up before these systems are tested in combat? Key steps include: clarifying who holds authority over AI systems across their lifecycle; building faster feedback loops between industry and the servicemembers actually using these tools; testing and red-teaming AI against battlefield conditions it never saw in training; and preparing servicemembers through professional military education to use AI with earned confidence rather than reflexive doubt.
Read the full report
About the author
Acknowledgments
The author thanks project advisers James Cartwright, retired Marine Corps general, Jack Shanahan, retired Air Force lieutenant general, and Douglas Small, retired Navy rear admiral, as well as all of those who participated in consultations to contribute to this report and its recommendations.
The Atlantic Council is grateful to Leidos, Primer AI, Latent AI, and Edge Case for their support of this report. This report was written and published in accordance with the Atlantic Council’s policy on intellectual independence, which requires all donors to agree to the council maintaining independent control of the content and conclusions of its work. The author is solely responsible for its analysis and recommendations.
Related reading
Explore the Program
Forward Defense leads the Atlantic Council’s US and global defense programming, developing actionable recommendations for the United States and its allies and partners to compete, innovate, and navigate the rapidly evolving character of warfare. Through its work on US defense policy and force design, the military applications of advanced technology, space security, strategic deterrence, and defense industrial revitalization, it informs the strategies, policies, and capabilities that the United States will need to deter, and, if necessary, prevail in major-power conflict.
Image: Several Low-cost Unmanned Combat Attack System (LUCAS) drones sit idle on tarmac at US Central Command (CENTCOM) base. Picture taken November 23, 2025. US Central Command Public Affairs.