Applied Intuition Wins $25M Navy AI Perception Deal

The US Navy has awarded Applied Intuition a $25 million contract to fit AI-powered sensor fusion to its warships, starting with five Arleigh Burke-class destroyers, Breaking Defense reported on Wednesday, citing the company. The deal could later extend “across the fleet”, according to the company.

Applied Intuition calls it a “first-of-its-kind production contract” in its press release. The system, called the Intelligent Perception System (IPS), will be deployed under what the company describes as the “DPRM Robotic and Autonomous Systems and PAE Mission Systems initiative”. The Navy has not published its own statement on the award, and the contract’s length has not been disclosed.

IPS fuses cameras and infrared into one picture for the bridge

According to the company, IPS takes feeds from shipboard cameras and infrared sensors, many of which were never designed to work together, and builds a single picture for watchstanders. It runs real-time detection, tracking and classification on Applied Intuition’s Acuity perception platform, and shows “automated alerts for vessels on dangerous trajectories”.

Breaking Defense connected that feature to the Navy’s own history. Sailors aboard the USS Fitzgerald and McCain could have used such a tool during their back-to-back collisions in 2017, it noted, which killed 17 sailors.

The release also says IPS keeps machine-readable records of surface contacts for pattern-of-life analysis, “replacing manual transcription from paper logs”. Those records can then be shared across the fleet.

The ships also collect training data and take updates at sea

The second job is data. IPS records what its sensors see and, according to Breaking Defense, sends it ashore “when bandwidth is available and security conditions permit”. A shore-based system then uses that data to train better models for manned and unmanned vessels.

Updated models go back to the ships over the air, “without requiring ships to return to port”, the company says. That could allow several refresh cycles within one deployment, rather than waiting for maintenance in harbour.

In a blog post, Applied Intuition describes this as its “Data Engine”. In that set-up, ships collect footage passively, the most useful clips are sent ashore by satellite, and improved models are pushed back to the fleet.

The contract follows a pilot called the Data Edge Collection Kit (DECK). According to Applied Intuition, DECK collected “thousands of hours of useful data” for model training. “DECK was massively successful,” Jason Brown, a general manager at the company, told Breaking Defense.

Applied Intuition began with software for self-driving cars, which remains a large part of its business. “The amount of data the Navy collects, and the operating environments it is constantly engaged in, [are] very similar to what we do gathering data to develop driverless cars,” Brown said.

The pipeline matters more than the alerts

The collision alerts will draw the headlines, but the bigger change is the data pipeline behind them. Militaries have plenty of sensor footage. Their problem is that it sits on hard drives until a ship returns to port, in formats nobody planned to train models on. If a destroyer can choose its useful clips, send them ashore through limited satellite links and receive a better model while still deployed, the Navy’s AI improves at the pace of software rather than refit cycles. That is the commercial self-driving playbook, and it helps explain why a car-software company won this contract rather than a traditional defence integrator.

The risks are in the parts the release passes over quickly. Sending data off a warship and accepting code updates at sea both widen the attack surface. A perception model that misreads a contact in a crowded strait is also a safety risk, not merely a bug. Breaking Defense’s line that data moves only when “security conditions permit” suggests the Navy knows this, but neither source sets out how updates are tested before they reach a ship.

At $25 million for five ships, this is still a modest deployment. The test is whether the Navy pays to extend it “across the fleet”. That fits a wider pattern of the US military buying commercial autonomy, from an unmanned surface vessel sailing with a carrier group to an MQ-25 controlled from 2,100 miles away. Investors are also backing start-ups such as Watchtower Labs, which raised $35m for military logistics AI, and the Navy has expanded its Gecko Robotics deal.

The earliest evidence that the system works will be the first over-the-air model update reaching one of those five destroyers while it is still at sea.