On October 8, 2026, the Futures Lab at the Center for Strategic and International Studies (CSIS) published an analysis finding that Chinese research vessels tend to stay near seabed infrastructure on the high seas longer than research vessels from other countries. Comparing the tracks of 223 vessels from 48 countries, the study found a statistical association between Chinese registry and the share of time a ship spends near seabed infrastructure. The report also examines in detail two Chinese research vessels that repeatedly followed survey-like tracks off Guam and in the western Pacific.
The analysis covers vessel tracks from January 1, 2020, to December 9, 2025. The findings were published as a white paper adapted for policy audiences from a manuscript submitted to an academic conference. General Atomics provided support for the research, and Optix data was used for vessel tracking. However, the fact that a vessel spent time near seabed infrastructure does not by itself prove sabotage or military intelligence gathering. The report shows that statistical analysis of tracks can help flag suspicious activity, but it also highlights how hard it is to determine a vessel's mission from its movements.
223 vessel tracks compared: Chinese ships show a link to time spent near seabed infrastructure
The research team selected fisheries research vessels, oceanographic research and survey ships, and similar vessels, then used the ship information service MarineTraffic to verify vessel types and Maritime Mobile Service Identity (MMSI) numbers. This yielded 223 vessels from 48 countries. It is a sample drawn under set criteria, not a complete census of the world's research vessels. For some vessels, no Automatic Identification System (AIS) data was available in certain years.
The metric used for comparison was the share of each vessel's estimated tracked time, calculated from time differences in AIS data, that was spent near seabed infrastructure on the high seas. The team defined "near" as within 1 km of the mapped route of a submarine optical cable or gas pipeline, to allow for activity in the surrounding area and positional error. This method shows only whether a ship was near mapped seabed infrastructure; it cannot directly confirm that a ship approached or touched the installation. Even a vessel classed as nearby did not necessarily know the exact position of the cable itself.
Based on footnote 2 on page 7 and pages 25–28 of the CSIS report, the main conditions and figures of the analysis are as follows.
| Figure / condition | Meaning |
|---|---|
| 223 vessels, 48 countries | Sample of research vessels selected after verifying vessel types and other details |
| Jan. 1, 2020 – Dec. 9, 2025 | Observation period of the track data |
| Within 1 km of seabed infrastructure | Distance used to classify a vessel as near a cable or pipeline |
| Estimated tracked time per vessel | Denominator used to calculate the share of time spent nearby |
| Adjusted R² of about 1.3% | Share of variation in the data that the regression model explains |
These figures describe the size of the sample, the spatial criterion, and the explanatory power of the statistical model. They do not indicate the number of sabotage incidents or the probability of an attack. In the regression by nationality, China, the United States, and Russia were treated individually, and the remaining 45 countries served as the common comparison baseline. The 1,177 observations listed in Appendix Table A3 are vessel-year records, not 1,177 vessels. Because data from multiple years for the same vessel are included, the number of vessels and the number of observations must be distinguished.
The analysis found that for most vessels, time spent near seabed infrastructure made up less than 1% of annual tracked time. In 2024, however, seven Chinese research vessels exceeded the 80th percentile, placing them in the top 20% or so of the sample. Still, U.S. research vessels also ranked near the top in some years, so lingering near seabed infrastructure is not behavior seen only among Chinese ships. Moreover, the comparison is of shares of tracked time, not a ranking of absolute time spent.
The statistical analysis counts time near gas pipelines as well as submarine optical cables. Alongside this analysis of seabed infrastructure as a whole, the report closely examines the tracks of two Chinese vessels that repeatedly conducted survey activity near specific telecommunications cables. The overall statistical trend and the case studies of individual cables should be considered separately.
Repeated survey-like tracks off Guam: what about mineral exploration?
The first case study is the Chinese oceanographic research ship Xiang Yang Hong 06. Between February and May 2024, it deployed three times to the high seas off Guam, leaving tracks that appear to be survey activity near the mapped routes of the submarine telecommunications cables Apricot and AAG. The tracks showed a "lawnmower" pattern of repeated passes along parallel lines, consistent with a navigation method commonly used for seabed mapping.
The research team focused on how these tracks related to the cable positions and to areas promising for mineral resources. The early survey areas were reportedly outside zones considered promising for cobalt-rich ferromanganese crusts and outside mineral-prospective areas identified by the U.S. Geological Survey (USGS). The tracks, meanwhile, were concentrated near undersea telecommunications cables. On this basis, the authors examined how well mineral exploration explains the initial activity.
The survey area in July included the mapped routes of the submarine cables ACC-1, AAG, and SEA-US. ACC-1, however, is a line still in the planning stage, so even if tracks overlap its planned route, that does not mean an existing cable was surveyed. In assessing the relationship between tracks and seabed infrastructure, operating facilities must be distinguished from planned routes.
From August to October, Xiang Yang Hong 06's survey range expanded and its track density rose. Activity extended into areas promising for mineral resources, and the authors suggest a shift from activity that may have been aimed at early information gathering to activity centered on mineral exploration. This is an inference from comparing tracks with seabed topography and the distribution of mineral resources, not a finding based on actual mission orders or collected data.
Nor does the fact that the early areas lay outside known mineral-prospective zones rule out scientific research. The report explains that data such as seabed topography, water temperature, salinity, and ocean currents can be used not only for oceanography but also for submarine operations and for evaluating sonar performance. In other words, information gathered through scientific research can have military value, but that alone does not establish that the purpose of the research was military. The track analysis does not reveal what instruments Xiang Yang Hong 06 actually used or what data it collected.
Ke Xue from the western Pacific to the Bering Sea, active near multiple cables
The second case is the Chinese oceanographic research ship Ke Xue. Between March and May 2024, it repeatedly deployed to waters west of Okinotorishima. CSIS identified tracks that ran near, or crossed the mapped routes of, the submarine telecommunications cables TGN-IA, Apricot, and JUPITER. Whereas Xiang Yang Hong 06 shifted its survey areas over time, Ke Xue is characterized by repeated deployments to the same waters followed by wide-ranging movement.
From mid-July, Ke Xue headed through the Sea of Japan and the Tsugaru Strait to the Bering Sea, leaving survey-like tracks near several international submarine telecommunications cables. The report also covers activity near the Russian submarine cable Polar Express. Some of the lines the tracks overlapped are ones in which China holds a stake, so it is difficult to generalize that the vessel chose to operate only near cables tied to other countries.
In evaluating such tracks, it matters not only whether a ship crossed a cable route but also whether it made repeated visits to particular waters and followed survey-like patterns across multiple areas. Still, repeated activity in the same place is not in itself evidence of sabotage. CSIS acknowledges that a single track or a cable crossing may have ordinary explanations, such as routine research activity. Port calls and AIS signal gaps likewise cannot by themselves indicate illicit activity.
The analysis of these two vessels shows that past tracks can be a useful clue for continuously monitoring activity around undersea telecommunications infrastructure. However, what AIS fundamentally reveals is a ship's position and movement history, not the work carried out underwater. Establishing what was actually measured and for what purpose requires corroboration from information beyond the tracks.
A statistical association, but no proof of any vessel's mission or intent
CSIS's statistical analysis found a positive association between being a Chinese-registered research vessel and the share of time spent near seabed infrastructure, and the association held in robustness checks that changed the analytical conditions. But confirming a statistical association is not the same as saying nationality adequately explains the behavior of individual vessels. According to Appendix Table A3, the regression model's adjusted R² is only about 1.3%, meaning the model explains only a small part of the variation in the share of time spent nearby.
Low explanatory power does not make the confirmed association meaningless. But because many factors besides nationality shape a vessel's behavior, there are limits to inferring a mission or activity from flag state alone. The authors note that future work should add behavior-related variables and expand the sample. They also report that estimates were less precise in additional models, so the results call for cautious interpretation.
CSIS expressly states that its findings do not establish the mission or intent of any individual vessel. A particular nationality or repeated tracks could be used as a cue for setting monitoring priorities, but they cannot by themselves support a finding of military intelligence gathering or sabotage. The behavioral characteristics visible in tracks must be clearly distinguished from speculation about the intent behind them.
In thinking about the security of undersea telecommunications infrastructure, it is also important to understand how communications would be affected if cables were actually damaged, in addition to assessing vessels' intent. A summary of a commissioned report released on September 22, 2026, by the U.S.-China Economic and Security Review Commission (USCC) notes that most undersea cable faults result from accidents or natural causes, and it analyzes scenarios for network impact when lines are damaged. This is a separate study from CSIS's track analysis, and it does not replicate that analysis.
According to the USCC analysis, damage to even a single submarine telecommunications cable could strain the available capacity on direct routes linking the United States with Taiwan, Japan, and the Philippines, shifting some traffic to slower detour routes. If multiple lines were damaged at once, spare capacity across the Pacific communications network could fall further short. However, this is a model analysis based on assumed failure conditions, not a measurement of communications outages that actually occurred. Nor does it predict that communications across all of Japan would stop.
To address such risks, the CSIS authors recommend stronger international information sharing and joint patrols, along with investment in regional repair hubs and multinational cable repair fleets. The USCC likewise calls not only for better monitoring but also for route redundancy and stronger repair capacity. These are policy recommendations, and no specific measures have been decided.
The study shows a statistical link between Chinese research vessels and time spent near seabed infrastructure, but it stops short of identifying any individual mission or intent. Going forward, whether investigators can confirm the survey work vessels actually performed and the data they obtained will be important in judging how valid track analysis is. It is equally essential to examine which communication routes would lose capacity if cables were damaged, and how far detour routes and repairs could make up for it. Preparing for both accidents and deliberate damage will require strengthening both the monitoring of suspicious activity and the ability to maintain and restore communications after a failure.