Navy approves AI strategy that puts deployment speed ahead of alignment risk

The Department of the Navy has approved a new data and artificial intelligence strategy intended to push AI into Navy and Marine Corps operations faster, including on ships and with expeditionary units. The department did not disclose a budget for the effort, but the plan is significant for defense technology suppliers because it frames slow adoption as a greater operational risk than some unresolved AI safety concerns.

Acting Secretary of the Navy Hung Cao signed the “Strategy to Weaponize Data and Artificial Intelligence,” putting it into effect immediately. The department said the document was developed over more than a year by its Chief Data and Artificial Intelligence Officer with AI specialists across the Navy and Marine Corps.

Cao said the strategy is meant to help the department “out-learn and out-fight any adversary” by using data and AI to improve decision-making. The department describes the end state as an “AI-first” fleet, a phrase that should be read as a procurement and operating signal rather than a technical specification.

Speed becomes the central metric

The strategy’s core framework is called the “Bits2Effects Cycle.” It covers the process of collecting military data, moving it, classifying it, analyzing it and turning it into operational decisions or actions. Feedback from those actions is supposed to update systems, tactics and training.

The department’s key measurement is “Mean Time to Effect,” or MTTE: the time between capturing new data and producing a military response or adaptation. According to the strategy paper, a force that shortens that loop over repeated cycles gains an advantage in a prolonged conflict.

The Navy announcement lists six goals for the plan:

• Move operational AI systems into use faster.
• Make data easier to find, access and use.
• Expand the technical infrastructure needed for AI deployment.
• Reduce approval delays.
• Improve data and AI skills among personnel.
• Work more closely with industry, academia, other government agencies and allies.

Many of the measures are expected by the first quarter of fiscal 2027, which ends in December 2026. By the end of fiscal 2029, the department says it aims to double the number of qualified data engineers, data scientists and AI and machine learning engineers.

Large models move closer to the edge

The plan calls for large language models and agentic AI systems to run aboard warships and alongside Marine Corps units, including when communications are degraded or unavailable. It also envisions service members building applications on top of those systems.

An “AI War Council” would rank use cases, allocate resources and pre-clear wartime changes to rules covering data sharing, classification and deployment. The strategy also adopts language from the Defense Department’s wider AI posture that says the danger of moving too slowly is greater than the danger of “imperfect alignment.” The document places that trade-off inside a “Wartime Approach,” meaning risk reviews and internal processes are to be handled with speed as a priority.

The Navy’s plan lands as generative AI use across the Defense Department is accelerating. Business Insider reported that GenAI.mil, the Pentagon’s central generative AI platform, reached 1.5 million daily users in June 2026, up from 80,000 at launch in December 2025. Reported uses range from office work to planning and combat support.

The Army is testing AI in a Next Generation Command and Control system for processing large volumes of data and supporting battlefield decisions. A Navy AI program reportedly reduced one submarine planning task from 160 hours to 10 minutes.

The strategy also points to a larger market shift. The Pentagon is increasing its use of commercial AI, including classified-network deployments and potential training of military-specific models on classified data. For AI vendors, the Navy’s document is another sign that defense customers will value deployability, security controls and edge operation as much as benchmark performance.

This story draws on original reporting from The Decoder.