A National Security Cutter stationed in the Eastern Pacific detects a small object floating in the waves, at half a mile. On a rolling ship at dusk, a human member of the watch crew could easily fail to notice it. The system doesn’t determine what action should then be taken; it merely informs a person that something might be worth examining.
A few years ago, such an alert wouldn’t have been available, but now it is one element of a much broader change since the U.S. Coast Guard, and the volunteer Auxiliary are gradually becoming some of the more interesting groups to adopt artificial intelligence within the federal government.
The change was no coincidence. In early this year, Rear Adm. Sean Regan, who is in charge of the Fourteenth Coast Guard District, stated that the service had to act swiftly in order to integrate data, AI, remote technology, space, and cyber capabilities if it was to keep up with its mission. The problem is not merely technical since it also relates to interoperability, training, and the modernization of the existing infrastructure.
A clear example of this is Ask Hamilton, a generative AI system which the Coast Guard has been developing over the last year. Since Alexander Hamilton, the founder usually regarded as the father of the service, is honored by the name, the tool is being constructed differently from a standard chatbot. Cmdr. Jonathan White, head of the Coast Guard’s cloud, data, and AI branch, has called it a system that has been trained on the service’s own records, manuals, and databases and that is designed to support the Coast Guard’s eleven statutory missions, ranging from search and rescue to port security.
White has stated that marine inspectors might be among the first to benefit, since AI could enable them to verify compliance findings in real time rather than having to take binders containing the relevant public law with them when they board each vessel. Similarly, naval engineers could make use of generative tools to compare ship blueprints with Coast Guard and Navy specifications much more quickly than a team working manually. According to White, the aim is not to have people send information to the AI, but rather to integrate the AI into the systems that they currently use.
That same philosophy is reflected in the Department of Homeland Security’s public list of Coast Guard AI applications. One of the systems uses camera-based object detection to identify people and life-saving equipment in the water, helping highlight potential search-and-rescue targets so a controller can decide where to allocate limited air and boat resources. A second system carries out part of the cybersecurity authorization process for new IT systems, a procedure which previously took weeks of staff time before a system could be deemed safe to operate. A third one aids with risk-based vessel inspections, enabling the service to direct its limited inspector manpower towards passenger vessels and cargo ships most likely to have safety issues. In all these instances, the final decision is left to a trained human.
The amount of money involved in this change is considerable. The Coast Guard intends to bid for a contract in the field of data, analytics, and AI support valued at over $100 million, a contract that will cover a range of mission and enterprise areas such as intelligence, logistics, finance, and acquisitions. The service has also asked for information on commercial AI tools that could be used to modernize procurement, a process that has long been criticized as slow and involves a lot of paperwork.
Congress has also backed the hardware aspect of the initiative by allocating $266 million for long-range unmanned aircraft and an additional $75 million for autonomous surface vessels through recent legislation. A portion of this capability is already in use. This spring, the Coast Guard’s Great Lakes District started launching autonomous sail drones fitted with radar, cameras, and collision-avoidance AI in order to assist with monitoring the lakes, gathering weather data, and keeping an eye on suspicious activity, with a human operator always on hand who can take manual control if needed. Autonomous V-BAT drones developed by Shield AI are also being deployed across the cutter fleet to increase the reach of a service that has always been expected to cover a large amount of water with a small number of personnel.
In this context, the Auxiliary fits with the plan. For many years, the Auxiliary has been developing its ability to operate unmanned aircraft, training and certifying members as remote pilots to carry out search patterns, waterway surveys, and disaster assessments alongside active-duty personnel. As the Coast Guard proceeds with AI-assisted detection and autonomous systems, the trained volunteer pilots and vessel examiners of the Auxiliary serve as a natural extension of this initiative by providing observation over the water and hands-on assistance at the docks in the same communities where these technologies are intended to operate.
A camera capable of identifying a person in the water is only useful if a trained crew is close enough to act; that has always been the Auxiliary’s specialty. Even if the technology improves signals, it still requires people to launch the boat, evaluate the situation, and carry out the mission.
Caution has not been spared for this change. White has been open that the service has been working to replace unauthorized and informal use of AI with approved and secure tools, and he has stressed the importance of guardrails and clear rules of engagement as adoption increases. Witnesses before Congress have expressed similar concerns regarding autonomous vessels, pointing out that present law still mandates the presence of human lookouts on Coast Guard vessels to avoid collisions, a requirement which has not kept up with the technology it is intended to regulate.
That tension between a service that wants to modernize and a legal and regulatory framework developed in an earlier era will probably influence much of the coming.
The reason why this situation is worth watching is that the Coast Guard is not pursuing the use of artificial intelligence for no particular reason; all the applications mentioned here, faster search-and-rescue cueing, quicker vessel inspections, streamlined cybersecurity approvals, and autonomous monitoring of remote waterways, achieve the same objective as that which the service and its Auxiliary have always strived for: getting the right person to the right place, with the right information, before a bad day turns into a tragic one.
A camera which detects an object in the water still requires a crew to launch the boat. A chatbot which replies to a policy question still depends on an inspector being willing to act on it. Innovation is not going to replace the people who provide service; it is instead giving them better sight and faster action, so that the watch they keep becomes more acute, quicker, and more accurate.
(The views expressed in this article are those of the author and do not reflect the official policy or position of the U.S. Coast Guard, the U.S. Coast Guard Auxiliary, the Department of Homeland Security, or any component thereof.)
