Invisibility is exactly the problem.
Beneath the ordinary bustle of airport security, a quieter crime has long moved undetected — the smuggling of marine life stripped from the world's oceans for soup, medicine, and underground markets. Scientists at Macquarie University have now trained an artificial intelligence to see what human eyes and older machines have missed, achieving 92% accuracy in identifying trafficked seahorses, shark fins, and sea cucumbers within airport X-ray scans. The work does not promise to end a billion-dollar trade, but it offers something meaningful: a way to make an invisible crime visible, using tools already present in airports around the world.
- Marine wildlife trafficking generates billions annually yet receives a fraction of the attention given to elephant or rhinoceros poaching, allowing vast quantities of sea creatures to cross borders undetected.
- Smugglers actively disguise their cargo — wrapping fins in foil, concealing seahorses inside children's toys — exploiting the sheer volume of airport luggage that no human team can fully inspect.
- Researchers trained a neural network on real trafficking seizures, deliberately introducing these concealment tactics into the training data to stress-test the algorithm against real-world conditions.
- The system achieved up to 96% accuracy for individual species and as low as 2% false positives, suggesting it could intercept shipments currently slipping through and generate evidence for prosecutions.
- The technology remains limited to three species, depends on three-dimensional CT scanners not universally available, and is designed to support — not supplant — human inspectors and detection dogs.
The illegal trade in marine wildlife moves quietly through the world's airports, tucked inside luggage alongside ordinary belongings. Shark fins, dried seahorses, sea cucumbers — these creatures are harvested and smuggled in volumes that rival more publicized wildlife crimes, yet they attract far less scrutiny. Scientists at Macquarie University, led by Dr. Vanessa Pirotta, set out to change that by turning a familiar piece of airport infrastructure into a new line of defense.
Their approach was practical by design. Rather than building new equipment, the team repurposed the three-dimensional CT scanners already used to detect explosives and biosecurity threats, training a neural network to recognize the distinctive shapes and densities of trafficked marine specimens. To make the training realistic, they sourced many of their 298 scans from actual seizures and deliberately complicated the images — hiding creatures inside toys, wrapping fins in foil, burying samples among legitimate goods.
The results were compelling: 95% accuracy for shark fins, 96% for seahorses, and 86% for sea cucumbers, with false positive rates as low as 1%. For sea cucumbers especially, where researchers suspect actual smuggling far exceeds recorded seizures, the tool could reveal a trade that has largely escaped notice.
Still, Pirotta and her team are measured in their claims. The algorithm covers only three species in a trade that spans hundreds. Many airports lack the advanced scanners the system requires. And no algorithm, however accurate, replaces the judgment of trained inspectors. The technology is one instrument in a larger effort — a way of giving border agencies a sharper eye for a crime the ocean's creatures have no means to resist on their own.
The ocean's black market operates in the shadows of legitimate commerce, hidden inside luggage and parcels that pass through airports every day. Shark fins destined for soup, seahorses dried for traditional medicine, sea cucumbers bound for underground markets—these creatures move across borders in quantities that dwarf the attention paid to more famous trafficking crimes. The illegal trade in marine wildlife generates billions of dollars annually, yet it remains largely invisible compared to the poaching of elephants or rhinoceroses. Now, scientists have developed a tool that might finally make this hidden crime visible.
Dr. Vanessa Pirotta and her team at Macquarie University created an artificial intelligence algorithm capable of identifying smuggled marine specimens with 92% accuracy, using technology already present in airports worldwide. The system works by training a neural network to recognize the distinctive signatures of trafficked species in three-dimensional X-ray images—the same CT scanners that detect explosives and biosecurity threats. Rather than inventing new infrastructure, the researchers repurposed equipment already in place, making the solution immediately practical for border security operations.
The team focused on three species: shark fins, seahorses, and sea cucumbers. Shark fins command high prices in certain food markets. Dried seahorses have been used in traditional medicine for centuries, creating steady demand that drives illegal collection. Sea cucumbers present a different problem—they are less frequently recorded in trafficking seizures, yet researchers suspect the actual smuggling volume far exceeds what authorities currently detect. To train their algorithm, the scientists gathered 298 scans from 68 total samples, many sourced from actual trafficking seizures. They created multiple scans of each specimen in different positions and orientations, then added realistic complications: fins wrapped in tin foil, creatures hidden inside children's toys, samples buried in luggage alongside legitimate goods. This mimicked the actual conditions smugglers employ.
The results were striking. The algorithm detected shark fins with 95% accuracy, seahorses with 96% accuracy, and sea cucumbers with 86% accuracy. False positives—instances where the system flagged something that wasn't actually smuggled goods—occurred in only 2% of shark fin cases, 1% of sea cucumber cases, and 9% of seahorse cases. These numbers suggest the technology could intercept shipments currently slipping through existing controls, disrupting trade routes and providing evidence for prosecutions.
Yet Pirotta and her colleagues are careful not to oversell their creation. The algorithm works only for three species, while hundreds of marine creatures face illegal trafficking. Not every airport possesses a three-dimensional CT scanner; many rely on older two-dimensional imaging equipment that cannot provide the detailed information the neural network requires. The technology will complement human inspectors and detection dogs, not replace them. As Pirotta noted, artificial intelligence is not a silver bullet. It is one tool among many, and its effectiveness depends on infrastructure, training, and the continued vigilance of people who understand that the ocean's creatures cannot protect themselves.
Notable Quotes
The trade of wildlife is cruel and unethical. For many, this may be the first people have heard of illegal trafficking of marine wildlife.— Dr. Vanessa Pirotta, Macquarie University
AI is not a silver bullet for detection, nor a replacement for human and sniffer dog detection.— Dr. Vanessa Pirotta
The Hearth Conversation Another angle on the story
Why does marine wildlife trafficking matter so much if we can't even see it happening?
Because invisibility is exactly the problem. A rhino poaching makes headlines. A container of dried seahorses passes through a port and nobody notices—but multiply that by thousands of shipments, and you've hollowed out entire populations. The damage is real; we just don't feel it the way we do with charismatic megafauna.
So the algorithm is meant to make the invisible visible?
Precisely. It turns an X-ray image—something a scanner already produces—into actionable intelligence. A customs officer sees a flag on a screen and knows to look closer. That's the whole game.
But you said it's not a replacement for human detection. Why not just let the AI do the work?
Because the AI only knows what it was trained on. Show it a species it's never seen, or a smuggling method nobody's documented yet, and it fails. Smugglers adapt. Humans adapt faster.
What about the false positives—the 9% error rate on seahorses?
Someone still has to check those. That's labor. That's cost. The algorithm saves time by narrowing the field, but it doesn't eliminate human judgment. It's a filter, not a replacement.
And the real barrier is that most airports don't have 3D scanners?
Exactly. This works beautifully in wealthy countries with modern infrastructure. But the trade flows through ports everywhere. You need the technology distributed globally, and that's expensive, and that's political.