The July 29 Pingtung demonstration paired Shield AI's Hivemind autonomy software with Thunder Tiger's SeaShark 600 and 800 in a coordinated surveillance and escort drill.
Two unmanned surface vessels ran a coordinated surveillance-and-escort drill off Pingtung, Taiwan, on July 29. The SeaShark 600 and SeaShark 800, both built by Taiwan's Thunder Tiger, planned their own waypoints, divided a search area, picked a target vessel out of the surrounding traffic, and handed the escort off from one hull to the other without a human choosing the handoff. The autonomy stack doing the work was Hivemind, from the U.S. defense-AI firm Shield AI.
Shield AI and Thunder Tiger described the drill as Hivemind's first multi-asset autonomous teaming on water, and the first time the software had been integrated with maritime radar and the Automatic Identification System, the transponder-based tracking layer that commercial ships use to broadcast their position. The mission combined onboard radar, electro-optical imagery, and AIS feeds into a single picture the autonomy stack could act on.
In Shield AI's framing, the software stack decouples autonomous behavior from the cost of any single platform. "Hivemind continues to scale mission autonomy across domains and platforms," said Nathan Michael in the joint release. One pilot now runs many hulls: a small, cheap USV becomes a distributed sensor and escort node rather than a one-off drone.
A fleet of small, fast boats with shared autonomy can be re-tasked mid-mission, pass targets between hulls to keep sensors on a contact, and absorb losses in a way that a single crewed patrol boat cannot. For an island defense ministry, that turns software into the limiting factor rather than hull count. If the capability lives in software, the unit of capability is a software update, not a new hull.
The procurement context lands the demo downstream of a real order, not in front of one. According to Taiwan News, Taiwan's Ministry of National Defense has placed a 1,300-unit order for the SeaShark class of unmanned surface vessels, and the platform has passed an NCSIST (National Chung-Shan Institute of Science and Technology) testing event off Yilan's Suao. The SeaShark 600, per the same reporting, carries 300 to 600 kilograms of payload, runs at up to 50 knots (93 km/h), and has a range of more than 350 kilometers from an aluminum-alloy hull that Thunder Tiger claims is roughly half the weight of a comparable fiberglass boat. Army Recognition's coverage of an earlier SeaShark 400 variant puts the smaller hull at 4,290 millimeters long, with a 43-knot (80 km/h) top speed, 150 to 250 kilograms of payload depending on whether the boat is running on batteries or gasoline, and a 150-kilometer range. A third hull, the SeaShark 800, is described in forum reporting as a deeper-draft, longer-range platform with a top speed above 50 knots and a range near 600 kilometers, though that figure has not been independently confirmed.
The demonstration builds on a strategic partnership Shield AI and Thunder Tiger announced earlier in 2026 to integrate Hivemind across Thunder Tiger's unmanned systems portfolio, covering both surface and aerial platforms including first-person-view systems approved under the Blue UAS program, a U.S. Defense Department effort to vet commercial drones for government buyers. The roadmap is broader than the Pingtung drill.
The Pingtung narrative rests entirely on the two companies' own announcement. No third-party reporting has confirmed the drill's mission outcomes, the sensor-to-decision latency, or whether the autonomous handoff behaved as described under degraded conditions. "The maritime domain is increasingly critical to Taiwan's security," the joint release said, and autonomous teaming "could expand distributed intelligence and surveillance," but those are vendor-stated stakes, not independent analysis. SeaShark 800 specifications come from a forum source and should be cross-checked against Thunder Tiger or Taiwan's defense ministry before being used as a load-bearing claim. The drill is not a weapons test and is not a deployment, and nothing in the source basis supports reading it as a change in deterrence.
The next test is whether the same software, on the same hulls, can be re-tasked quickly enough to matter operationally. The 1,300-unit Taiwan MoD order, if it ships on the schedule Taiwan News describes, will give that question a real fleet to ask it on.