The US Department of the Navy has formally approved a strategy to make AI a core part of how it fights. Signed by Acting Secretary of the Navy Hung Cao, it took effect immediately. The strategy was over a year in development and covers both the Navy and the Marine Corps.
The core logic is straightforward. In modern conflict, the side that processes new information and adapts fastest wins. The Navy watched Ukraine show this in real time: on that battlefield, one side can deploy a new tactic and the other side has a countermeasure ready within 24 hours. The strategy wants the Navy to operate at that speed.
To do that, the Navy laid out a five-stage cycle it calls Bits2Effects. Data gets collected from sensors and weapons automatically, transported even when communications are jammed, classified with less human delay, analyzed by AI, and then fed directly into decisions and tactics. Lessons from each cycle feed back into the next one. The key metric is how long this entire loop takes. Shorter is better.
The strategy calls for running large AI models directly on warships and with Marine units in the field. These systems need to work even when cut off from central networks. Sailors and Marines would build their own tools on top of them. A new body called the AI War Council would decide which AI projects get priority and pre-approve rule changes needed during wartime.
The timelines are concrete. Most measures are supposed to be in place by December 2026. The number of qualified data engineers and AI specialists across the Navy and Marine Corps is set to double by the end of 2029.
This strategy did not emerge in a vacuum. The Pentagon's central AI platform, GenAI.mil, went from 80,000 daily users when it launched in December 2025 to 1.5 million by June 2026, representing nearly half of the entire Department of Defense workforce. Pentagon staff have already built over 100,000 semi-automated AI tools using that platform in under five weeks. One Navy recruiting office reportedly cut a multi-year database project down to three months using AI. A separate Navy program reportedly cut a submarine planning task from 160 hours to ten minutes.
The strategy also reflects a deliberate political choice: moving slowly is treated as a bigger risk than deploying AI that is not perfect. That is not just a line in a document. It shaped what happened with Anthropic. The company refused to allow its AI model, Claude, to be used for fully autonomous weapons or mass domestic surveillance. The Trump administration responded by banning federal agencies from using Anthropic's products and designating the company a supply-chain risk, a label usually reserved for foreign adversaries. Hours later, OpenAI signed a classified deal with the Pentagon to fill the gap. The Pentagon has since signed AI agreements with seven companies, including Google, Microsoft, and Nvidia.
On the other side, China is not sitting still. Researchers at Georgetown University analyzed nearly 3,000 procurement contracts from China's military published between 2023 and 2024. The documents show the military testing AI for drone swarms, satellite warfare, ship tracking, deepfake-powered disinformation, and cyber operations. Most contracts have short development timelines of three to six months, meaning China is running rapid, low-cost experiments rather than long, expensive programs. China's 15th Five-Year Plan, approved in early 2026, locks this approach in through 2030.
The US still holds a significant lead overall: more computing power, more trained AI talent, and more operational experience. But the Georgetown researchers flag one genuine risk. Washington publicly humiliated one of its leading AI suppliers, Anthropic, during a period when China is actively trying to close the gap. Expanding these partnerships matters more for national security than winning political arguments over contract language.
For anyone running a business that works with defense contractors, government agencies, or large regulated industries, the pattern here is worth noting. The US military just told the world that it will treat AI adoption speed as a survival requirement and will remove suppliers that slow it down. That posture is spreading. Private sector procurement, insurance, logistics, and operations teams will face the same pressure from their own leadership sooner than most expect.