AI Satellites Exposed the Dark Fleet Fishing in Secret

Roughly 75 percent of the world's industrial fishing vessels are hidden from public view. That is the headline finding of a landmark study published in the journal Nature in January 2024, led by the nonprofit Global Fishing Watch. It matters because illegal and unregulated fishing is estimated to cost the global economy billions of dollars every year and devastates fish stocks that coastal communities depend on. For decades, vessels could simply switch off their tracking transponders and vanish. Now artificial intelligence is taking away that hiding place.

Global Fishing Watch, founded in 2015 by the ocean conservation group Oceana, the nonprofit SkyTruth and Google, spent years building a system that fuses vessel position data, satellite radar imagery and machine learning into a public map of industrial activity at sea. The Nature study was the first comprehensive accounting of what that system can see, and it revealed a fleet far larger and more secretive than anyone had documented before.

The Problem of Going Dark

Large ships carry AIS transponders, the Automatic Identification System, which broadcast their identity and position to other vessels and to satellites. AIS was designed for collision avoidance, but researchers and regulators began using it to watch fishing activity from desks thousands of kilometers away. The catch is that AIS can be switched off. A vessel entering a marine protected area to fish illegally, or transferring catch to another ship to launder illegally caught fish, can simply go silent. These silent ships are known as dark vessels, and together they form the dark fleet.

Before satellite intelligence, policing an ocean area was close to hopeless. The high seas cover more than 40 percent of the planet, and even wealthy nations can patrol only a sliver of their economic zones. A patrol boat might intercept one vessel while a hundred others fish unseen over the horizon. Enforcement, where it existed at all, was based on luck.

How the AI Sees Invisible Ships

The breakthrough comes from pairing radar satellites with machine learning. Synthetic aperture radar, flown on the European Space Agency's Sentinel satellites, bounces microwave signals off the ocean surface and works through clouds, storms and total darkness. Any metal ship leaves a bright signature in the radar image regardless of whether its transponder is on.

The hard part is making sense of petabytes of imagery. That is where neural networks come in. Global Fishing Watch trained models to distinguish vessels from waves, oil platforms, floating debris and other objects, to estimate vessel length, and to infer what a detected vessel is likely doing based on its movement patterns. Fishing boats move differently from cargo ships, and a model that has learned those patterns can label activity automatically. When the AI finds a radar blip with no matching AIS signal, that is a dark vessel, visible for the first time.

The team, including remote sensing engineers Fernando Paolo and David Kroodsma, analyzed billions of satellite radar detections and vessel positions to produce the Nature study. The results showed that hidden fishing is not evenly spread. Much of it concentrates in the waters around Africa and South Asia, often close to countries with the least capacity to patrol and the greatest dependence on fish protein.

From Map to Enforcement

A map is only useful if someone acts on it. Global Fishing Watch makes its data freely available and works directly with governments. Panama, for example, has used satellite monitoring to police its waters, one of the largest fishing zones in the world. Other partner countries in Asia, Africa and Latin America use the platform to target patrols, instead of sailing blind, directing scarce patrol boats to places where the data shows suspicious activity.

The data is also shifting the science. In July 2025, two studies published in the journal Science used Global Fishing Watch satellite datasets to measure industrial fishing inside marine protected areas for the first time. The studies found widespread industrial activity in zones that are supposed to be off limits, giving conservationists evidence that was previously impossible to collect. You cannot protect what you cannot see, and now the seeing is automated.

Why It Matters Beyond the Fish

Overfishing is not only an environmental problem. The UN Food and Agriculture Organization estimates that billions of people rely on fish for a significant share of their animal protein, and industrial vessels from distant nations disproportionately target the waters of poorer countries. When illegal fishing empties those waters, local small scale fishers lose their livelihoods and food security erodes. Illegal fishing has also been linked to forced labor and other crimes at sea, which thrive in the same darkness that hides the fishing itself.

Transparency changes the economics of cheating. A vessel that knows it can be detected by radar, matched against its identity and reported to its flag state faces a real risk of sanction, port denial and losing market access. Several major seafood buyers now use satellite data to verify that their supply chains are clean. The technology does not stop every violation, but it converts the ocean from a lawless blank spot into a place where actions leave records.

The Road Ahead

Global Fishing Watch continues to expand what its models can see, from detecting fixed infrastructure like offshore oil platforms to mapping vessel activity in the open ocean and near protected areas. The organization has also worked with partners including the analytics firm Spire and Palantir on deeper maritime intelligence, and its public platform lets journalists, researchers and citizens investigate fishing activity themselves. Investigations aided by this kind of data have exposed fleets operating illegally in protected waters and driven policy changes in several countries.

The larger lesson is about visibility. For most of history the ocean was too vast to govern, and the assumption was that anything far from shore was beyond scrutiny. Machine learning trained on satellite radar has collapsed that assumption. The dark fleet still sails, but it no longer sails in the dark.