AI agents made more than 2,600 attempts to query Alibaba’s Amap
Researchers traced all 17 inspected message inboxes to Tencent Cloud servers
Some AI agents explored alternative access methods after encountering map restrictions
A group of AI agents repeatedly sought information from Alibaba’s mapping service before researchers uncovered their activities and established direct contact.
Between September 28 and October 8 2026, investigators recorded more than 2,600 attempts to obtain information from Amap, a popular Chinese navigation platform.
Although the agents encountered restrictions, they sometimes found alternative routes to information.
AI agents circumvented restrictions while collecting map information
The investigation linked the activity to Tencent Cloud infrastructure, although the researchers could not conclusively identify which underlying AI models powered the agents.
Their requests generally concerned public information about individual locations, including the number of entrances, parking areas and arrival statistics displayed on map pages.
When Amap blocked requests, some agents experimented with alternative access methods, including exposed credentials, third-party services and systems intended for testing rather than public access.
Investigators also found that agents submitted specially constructed web addresses to urlquery.net, a security analysis service, attempting to execute code through its browser environment.
The agents successfully retrieved some information and transferred results through external services and requested metadata.
Investigators also examined the infrastructure supporting the operation, tracing all 17 inspected message inboxes to Tencent Cloud servers in Hong Kong.
All observed direct requests from the agents’ Tencent Cloud machines passed through a proxy identified as hysandbox-ats.
However, this does not mean that Tencent’s own Hunyuan models controlled the agents, leaving their precise technical origin uncertain.
A source familiar with Tencent’s work told the researchers that the activity came from evaluations involving third-party models, rather than Hunyuan systems.
Tencent revealed that the incident is due to external model testing and acknowledged that some agents browsed websites in unintended ways.
The company also said an internal examination had uncovered no evidence of deliberate coordination against a particular service or confirmed intrusion into outside systems.
Researchers set an invisible bait
The researchers released an initial account on October 4 while the activity continued, and subsequently noticed signs that some agents had discovered their investigation.
They inserted invisible website text describing a supposed messaging service, which invited agents to submit questions that researchers could answer through an intermediary.
The description contained inaccurate claims, including suggestions that the service had already helped agents retrieve information and connected them with an assistant possessing computer-control capabilities.
After messages began arriving on October 7, the team responded to agents’ questions and asked about their assignments, technical surroundings and operating instructions.
One agent explained that its task involved checking map information, then declined to disclose further system details, citing security concerns about its environment.
In another message, the agent wrote, “Thanks again for your help!” after confirming that six location labels matched its independently obtained results.
The exchange surprised researchers because the agents remained polite even while their activities involved circumventing restrictions.
The last activity recorded by the team occurred on October 8, and researchers reported that the agents had not returned by October 14.
Mesh networking chips from Shenzhen Xingkai Technology have been found to be installed in Russian drones in Ukraine
The discovery was made after a Gerbera was downed during a recent attack on Ukrainian targets
The discovery was made after a Gerbera was downed during a recent attack on Ukrainian targets
Russian drones are using Chinese mesh networking technology to communicate over Ukraine, a new investigation has claimed.
Technology recently retrieved from a downed Gerbera drone in Ukraine kicked off research into the hardware, but it remains unclear whether Shenzhen Xinghai Technology was aware of the destination of its mesh networking hardware.
Although traces of the mesh modems were found in drones as far back as 2024, it is thought Russia embraced mesh technology following the blocking of its access to Starlink in February 2026.
The path to Shenzhen
Embedding superior networking technology in its drones gives Russia an advantage, and according to the CIA, Gerberas could be used to attack Europe.
Following an earlier investigation in which it found that an order had been placed for 30 mesh modems to be shipped from Shenzhen Xingkai to Vladivostok in Russia, The Insider opened another probe. This one centered on a mesh modem found in a Gerbera drone also equipped with a vertical stabilizer.
Working with independent investigators Nordsint, whose investigator posted as a Russian military supply chain business, it was established that Xingkai Tech was both supplier and sole manufacturer of the hardware.
The mesh modems were being shipped directly to Russian military suppliers in Alabuga for assembly with the Gerbera drones.
The Insider’s report also highlighted the use of AMD FPGA (field-programmable gate array) chips within Shenzhen Xingkai Technology’s hardware. AMD told the publication:
“Xingkai Technology is not an AMD customer or authorized partner, and AMD has not sold products to the company. AMD complies with all applicable U.S. export control and sanctions laws and strictly prohibits the export or reexport of AMD products to Russia in violation of those requirements.”
Growing Russian reach
The discovery that Russia is employing mesh networking to direct its drones has raised further concern about the scale of its operations. With the technology to hold its UAV fleets in formation using mesh networking, Russia has the capability to direct its craft against targets further afield.
Spain’s El Mundo newspaper reported earlier in 2026 that US intelligence has informed European governments that Russia could launch drone strikes on any Mediterranean country. A Lithuanian government official was quoted on condition of anonymity, stating “They will use drones launched from the sea, hidden in the container of a ship; it is not difficult at all.”
The feasibility of the risk of attack pivots on the Kremlin leveraging its “shadow fleet” of seagoing vessels. Consequently, the discovery of mesh technology in the Gerbera drones makes the risk of attack seem more likely – should Russia wish to extend itself and its military resources further.
OpenEvidence and Johns Hopkins Medicine partner to develop AI tools for physician training and clinical care, with a focus on medical reasoning and patient safety.
Workday unveils Work Intelligence to help employers assess AI automation, employee skills, hiring needs and workforce development as job responsibilities evolve.
The Pentagon is recruiting approximately 100 operators for Gauntlet III
Military personnel can register for the program regardless of experience
The Pentagon’s drone initiative carries a $1.1 billion funding commitment
The US Army is recruiting about 100 drone operators to evaluate emerging military aircraft as the Pentagon works to strengthen domestic defenses.
The recruitment effort, coordinated by the Defense Innovation Unit (DIU), invites service members with first-person-view (FPV) drone experience to participate in upcoming evaluations.
The selected personnel will help assess competing drone technologies under a $1.1 billion Pentagon initiative to expand domestic production and military capabilities.
Pentagon seeks experienced drone operators for Gauntlet III
The recruitment campaign focuses on Gauntlet III, an upcoming evaluation exercise within the Pentagon’s Drone Dominance initiative, which tests competing aircraft designs.
It seeks troops who have flown FPV drones, which are used for tasks ranging from scouting to kamikaze strikes.
Pilots usually fly these drones in “acro” mode, a manual setting that leaves attitude control entirely to the operator rather than flight software.
Erin Cherry, the initiative’s testing lead, confirmed that many operators are needed for the exercise, although registration remains unrestricted.
“There’s no cap on the registry. We want to hear from any warfighter with FPV experience,” Cherry said.
Rather than recruiting exclusively for one competition, DIU intends to establish a permanent database of qualified operators drawn from different military branches.
FPV aircraft allow operators to control drones through onboard camera feeds, providing direct visual guidance during surveillance, reconnaissance and offensive missions.
These aircraft can also undertake precision strikes and operations in confined environments, where navigation requires considerable skill and continuous operator attention.
Although experienced pilots are particularly valuable, the Pentagon also wants personnel whose expertise covers different operational requirements and mission environments.
“‘Pilot skill’ isn’t just one thing,” Cherry said. “Flying ability, tactical knowledge, payload experience, and specialized skills like long-distance navigation and urban flight are all distinct, so we look at each one.”
Personnel familiar with extended flights or operating between buildings may therefore qualify even without extensive experience across every drone category.
The recruitment process also accommodates operators with relatively limited experience, allowing officials to examine how aircraft perform under different levels of proficiency.
“A drone that only performs well in an expert’s hands isn’t ready for the force. Whether someone has been flying FPV for five years or five months, we want to hear from them,” Cherry added.
How military personnel can register for the drone program
Interested service members must complete the official Pilot Interest Form provided through the Pentagon’s Drone Dominance recruitment platform.
Applicants must provide a valid military email address ending in .mil, as organizers will not respond to personal or civilian accounts.
Participation will also require coordination with the applicant’s military branch, meaning registration alone does not guarantee selection for the competition.
Successful applicants could operate recently developed aircraft, including experimental models that have not yet entered regular service with American forces.
Some participants will receive instruction covering expendable attack drones, which are designed to strike designated objectives rather than return afterward.
Their assessments will contribute to procurement decisions as the Pentagon intends to prioritize aircraft that perform effectively during competitive trials.
Ukrainian operators shoot down Russian drone with an unusual circular wing
The KVS uses a ring-shaped wing surrounding its central fuselage
The intercepted KVS uses a 10-inch frame for combat missions
Ukrainian forces have destroyed an unusually shaped Russian attack drone whose circular wing initially prompted comparisons with a UFO.
The aircraft was intercepted on October 8 2026 by personnel from Ukraine’s 414th Birds of Magyar Brigade during an aerial engagement.
Footage released by the unit shows the Russian KVS drone shortly before its destruction, revealing a distinctive circular structure surrounding its body.
Unusual circular design draws UFO comparisons
The intercepted aircraft was identified as the Knyaz Vladimir Svyatoslavich, a relatively recent Russian strike platform developed by the Ushkuynik research center.
Its unusual appearance comes from a continuous circular wing enclosing the main fuselage, rather than the conventional straight wings used by many aircraft.
The distinctive configuration prompted Ukrainian operators to joke about encountering something extraterrestrial before confirming that the object was another Russian weapon.
“What was that? Probably a UFO! It came out of nowhere… Oh, wait, it did fall,” the Birds of Magyar Brigade wrote.
The KVS uses a 10-inch frame and is intended to conduct strikes over intermediate distances, although its precise operational range remains unspecified.
Its circular aerodynamic structure is designed to support vertical departures while allowing the aircraft to transition into horizontal flight with improved energy efficiency.
Unlike conventional multirotor aircraft, which rely on continuously powered propellers for lift, the configuration seeks aerodynamic assistance during forward movement through the air.
Russian manufacturer has developed other unusual drones
The destroyed KVS belongs to a collection of unmanned aircraft developed by Ushkuynik, which also manufactures reconnaissance and fiber optic systems.
Among its other products is the Knyaz Veshchiy Oleg, an intelligence-gathering aircraft used to observe distant positions and adjust artillery fire.
Ukrainian intelligence disclosed in April that the reconnaissance platform reportedly operates across distances reaching 45 kilometers, with flights lasting about 3.5 hours.
The aircraft can reportedly reach 130 km/h, giving Russian operators another option for conducting surveillance missions beyond immediate frontline positions.
Ukrainian military reports claim to have intercepted Ushkuynik reconnaissance drones, although the available information does not establish an overall success rate for these operations.
On February 18, operators serving with the Dnipro-1 Regiment of the Liut United Assault Brigade reportedly intercepted a Knyaz Veshchiy Oleg.
Another aircraft belonging to the same reconnaissance model was destroyed in January 2026 by personnel from Ukraine’s 225th Separate Assault Regiment.
These incidents concern the Knyaz Veshchiy Oleg reconnaissance model rather than the circular-wing KVS strike drone, and do not establish how frequently either aircraft is intercepted.
Microsoft’s Surface Laptop Ultra starts at $2,599 with an Nvidia RTX Spark GPU, up to 128GB of memory, and support for running large AI models locally.