AI is not a silver bullet for detection, nor a replacement for human expertise
Every day, the ocean's most vulnerable creatures pass through airport security in silence — tucked into suitcases, wrapped in cloth, hidden among ordinary belongings. A team of researchers at Macquarie University has answered this quiet crisis with an AI algorithm capable of identifying smuggled shark fins, seahorses, and sea cucumbers inside luggage with 92% accuracy, using CT scanners already present in many airports. The work addresses not only the difficulty of detection but the deeper invisibility of marine trafficking — a billion-dollar trade that rarely commands the moral urgency of its terrestrial counterparts, even as it erodes the ecosystems that sustain ocean life.
- Marine wildlife smuggling moves in silence — shark fins, dried seahorses, and sea cucumbers slip through airport checkpoints daily as part of a billion-dollar illegal trade that rarely triggers alarm.
- The true scale of the crisis is itself hidden: sea cucumbers, for instance, are suspected to be smuggled far more than seizure records reveal, leaving authorities unable to fully grasp what they're fighting.
- Researchers trained a neural network on CT scan images of actual trafficked specimens — testing it against items wrapped in foil, buried in clothing, and concealed inside children's toys to mirror real smuggler tactics.
- The algorithm achieved 95% accuracy on shark fins, 96% on seahorses, and 86% on sea cucumbers, offering a credible new layer of detection without requiring airports to overhaul existing infrastructure.
- False positives still demand human review, many airports lack 3D CT scanners, and the system covers only three of many trafficked species — making it a promising tool, not a complete solution.
The smuggling of marine life across borders happens quietly. Shark fins tucked into suitcases, dried seahorses wrapped in cloth, sea cucumbers packed among clothes — these contraband items slip through airport checkpoints every day, part of an illegal trade worth billions annually. Unlike the trafficking of rhino horns or elephant ivory, the poaching of ocean creatures receives far less public attention, even as it pushes vulnerable species toward extinction.
Dr. Vanessa Pirotta and her team at Macquarie University set out to change that. Using CT scanners already installed at many airports, they trained an AI algorithm to recognize three commonly smuggled marine species — shark fins, seahorses, and sea cucumbers — with 92% overall accuracy. The system analyzes three-dimensional X-ray images, identifying the distinctive shapes and densities of these creatures even when concealed in foil, clothing, or children's toys.
The dataset was modest but deliberate: 298 scans drawn from actual wildlife trafficking seizures, tested against scenarios designed to mimic real smuggler tactics. Results were striking — 95% accuracy for shark fins, 96% for seahorses, and 86% for sea cucumbers. False positives varied by species but remained low for shark fins and sea cucumbers, suggesting the tool could intercept shipments currently evading notice.
Pirotta is careful about the algorithm's limits. It covers only three species among many that are trafficked. Every false positive still requires a human inspector. And many airports rely on older 2D imaging technology, not the 3D CT scanners the system depends on. The algorithm is designed to complement human expertise, sniffer dogs, and traditional screening — not replace them. Artificial intelligence, as Pirotta frames it, is one instrument in a larger effort to make the ocean's invisible trade visible before it reaches its destination.
The smuggling of marine life across borders happens quietly, hidden inside luggage that passes through airport security without raising an alarm. A shark fin tucked into a suitcase, dried seahorses wrapped in cloth, sea cucumbers packed among clothes — these are the contraband items that slip through checkpoints every day, part of an illegal trade worth billions of dollars annually. Unlike the high-profile trafficking of rhino horns or elephant ivory, the poaching and sale of ocean creatures receives far less attention, even as it destabilizes fragile marine ecosystems and pushes vulnerable species toward extinction.
A team of scientists led by Dr. Vanessa Pirotta at Macquarie University has developed a tool to change that. Working with existing X-ray CT scanners already installed at many airports, they trained an artificial intelligence algorithm to recognize three commonly smuggled marine species — shark fins, seahorses, and sea cucumbers — with 92% accuracy. The system works by analyzing three-dimensional images created from multiple X-rays of a single object, allowing the algorithm to spot the distinctive shapes and densities of these creatures even when they're wrapped, hidden in toys, or buried among other items in a bag.
The research began with a straightforward problem: detecting trafficking in progress is extraordinarily difficult, which means authorities struggle both to stop the trade and to understand its true scale. Shark fins command high prices in food markets. Dried seahorses are valued in traditional medicine. Sea cucumbers are less frequently recorded in seizures, but researchers suspect they're smuggled far more often than current detection rates reveal — the animals are illegally overfished, and the true extent of the trade remains hidden. To build their detection system, the scientists gathered samples from wildlife trafficking seizures and created 298 scans total: 20 sea cucumber samples, 30 seahorse samples, and 18 shark fin samples. Each sample was scanned multiple times in different positions and contexts, and the researchers deliberately tested their algorithm against images showing the items wrapped in foil, concealed in clothing, or tucked inside children's toys — mimicking the actual tactics smugglers use.
When tested on images the algorithm had never encountered before, the results were striking. The system detected shark fins with 95% accuracy, seahorses with 96% accuracy, and sea cucumbers with 86% accuracy. False positives — instances where the algorithm flagged something that wasn't actually contraband — occurred in 13% of cases overall, though this varied by species: just 2% for shark fins, 1% for sea cucumbers, and 9% for seahorses. These numbers suggest the technology could intercept shipments currently evading detection and help authorities build cases against traffickers.
Yet Pirotta is careful not to oversell what the algorithm can do. Many other marine species are trafficked beyond these three, and every false positive still requires manual verification by a human inspector. Not every airport has access to the expensive 3D CT scanners the system relies on; many facilities still use older 2D imaging technology. The algorithm is a complement to existing detection methods — human expertise, sniffer dogs, and traditional screening — not a replacement for them. As Pirotta notes, artificial intelligence is not a silver bullet. It is one tool among many, and its effectiveness depends on how it's integrated into the broader work of catching wildlife crime before it reaches its destination.
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 get so little attention compared to, say, elephant ivory or rhino horns?
It's partly invisibility. A shark fin or a dried seahorse doesn't have the same symbolic weight in the public imagination. But the damage is just as real — you're disrupting entire ocean food webs, pushing species toward extinction, and the scale is staggering. Billions of dollars a year.
So the AI algorithm is the solution?
It's part of one. The algorithm is genuinely useful — 92% accuracy is high enough to flag bags for human inspection. But Pirotta is explicit about this: it's not a silver bullet. You still need people to verify the flags, and you need the expensive scanners in the first place.
What about the false positives? Nine percent for seahorses seems high.
It does, and that's the friction point. Every false positive means an inspector has to stop and look at something that isn't actually contraband. That's time and labor. But it's also better than the alternative — letting the real smuggling through.
Why these three species specifically?
They're commonly trafficked and they have distinct enough shapes that the algorithm can learn to recognize them in scans. But there are hundreds of other marine species being illegally traded. This is a proof of concept, not a complete solution.
What happens next?
The real test is deployment. Does this actually work in live airport settings? And can the approach be adapted for other species? Right now it's a laboratory success. Making it operational at scale is a different challenge entirely.