The PLAAF reveals an AI system capable of coordinating over 100 aerial units, a major advancement that remains dependent on data and networks.
In Summary
In early August 2026, China revealed an AI strike planning system designed for the People’s Liberation Army Air Force. According to CCTV, this tool can handle exercises involving hundreds of targets, dozens of formations, and more than 100 tactical units. It assists headquarters in prioritizing objectives, allocating weapons, organizing routes, and synchronizing multiple attack waves. The novelty lies not in an autonomous weapon, but in the partial automation of planning tasks. The system is also intended to facilitate collaborative operations among piloted aircraft, drones, bombers, airborne early warning aircraft, and missiles. This evolution aligns with Chinese ambitions surrounding the J-20S and the GJ-11. However, it remains exposed to several vulnerabilities. A military artificial intelligence depends on data quality, network availability, and its resilience against jamming, decoys, and cyberattacks.
Public Unveiling Marks a Milestone in Chinese Air Warfare
The revelation was widely reported on August 3, 2026, following the broadcast the previous day of an episode of the documentary series Zhisheng by state broadcaster CCTV. The series accompanied the 99th anniversary of the People’s Liberation Army.
The footage featured an intelligent strike preparation system developed by a team led by Senior Colonel Deng Jianping at a base of the People’s Liberation Army Air Force, or PLAAF.
According to the official account, the system passed a final live-fire evaluation. Missiles and bombs reportedly struck their assigned targets. The system was then employed during an exercise involving over 100 tactical units. Chinese media also claim it has already been used in several PLAAF missions, without specifying their nature, date, or level of complexity.
This messaging must be read with caution. Beijing has published neither the official full name of the program, nor its IT architecture, nor its performance compared to a staff working without artificial intelligence. No error rate, average planning delay, or detailed results have been made public.
The demonstration nevertheless provides important information. China is no longer limiting military artificial intelligence to automatic image recognition or drone piloting. It is seeking to insert it at the air campaign level, where objectives, aircraft, missiles, and attack windows are allocated.
The System Transforms an Operational Puzzle into a Computing Problem
A massive airstrike does not consist of simply launching a large number of aircraft simultaneously. It requires solving a set of often conflicting constraints.
It is necessary to identify the most critical targets, select appropriate weapons, and determine the required quantity of munitions. Aircraft range, fuel, sensors, and radar cross-sections must be taken into account. Adversary defenses, weather, penetration routes, aerial refueling, and collision risks must also be integrated.
The problem quickly becomes too vast to process manually. With several hundred targets and dozens of formations, potential combinations run into the millions. A change in the situation can render part of a plan obsolete in minutes.
The value of artificial intelligence here lies in its ability to rapidly compare a large number of solutions. It can propose force allocation, rule out incompatible options, and recalculate the plan when an aircraft, tanker, or target is no longer available.
Target Selection Becomes Optimization Under Constraints
The primary role of the system is to prioritize objectives. Not all targets have equal value, and not all need to be struck immediately.
A command center, an air defense radar, an airfield runway, and a fuel depot produce different military effects. Destroying a radar can open a corridor for subsequent aircraft. Attacking a runway can prevent enemy fighters from taking off. Striking a communications node can disrupt multiple units at once.
The software must therefore consider the relationships between objectives. It does not merely seek to destroy a list of targets; it must produce a cumulative effect on the adversary’s posture.
The tool can then match each target with a strike asset. A cruise missile may be used against a heavily defended fixed objective. A guided bomb might suit a closer structure. A fighter may be assigned to escort a bomber rather than conduct a strike mission.
The intended result is better resource utilization. Employing an overly expensive weapon against a secondary target wastes capacity. Using inadequate munitions forces a follow-up strike, giving the adversary time to react.
Wave Coordination Aims to Shorten the Strike Cycle
An air campaign is generally organized into multiple waves. Initial missions seek out radars, command posts, and antiaircraft batteries. Subsequent waves attack primary targets. Other aircraft provide escort, jamming, refueling, or damage assessment.
These operations must be precisely synchronized. A jammer arriving too early reveals an imminent attack. An anti-radiation missile launched too late fails to protect the formation. A bomber penetrating before defenses are neutralized exposes itself unnecessarily.
The Chinese system seeks to automate part of this synchronization. It must determine launch times, strike order, and intervals between different formations. It must also recalculate the plan as new information arrives.
This function directly aims to shorten the kill chain. This chain connects detection, identification, decision, engagement, and damage assessment. The faster this loop runs, the less time the adversary has to relocate forces, turn off radars, or disperse aircraft.
The Platform Does Not Command Weapons Alone
The term artificial intelligence can create a false impression. Nothing in the public information indicates that the Chinese system can authorize a strike on its own or independently command the release of a bomb.
Based on currently available information, it functions as a decision-support tool. It prepares options and recommends asset allocation. Command personnel retain responsibility for the plan and the engagement of weapons.
This distinction is vital. Planning software can be highly automated without being an autonomous weapon. It can compute a route, compare several munitions, or propose a priority order while leaving the final decision to a human being.
The technology employed has not been detailed. Nothing proves that the system relies on a large language model comparable to civilian generative artificial intelligence tools. It may combine classical operational research methods, expert-written rules, simulations, optimization algorithms, and certain machine learning models.
Senior Colonel Deng Jianping’s remark is revealing. According to him, the main difficulty was not writing the code, but understanding the logic of the battlefield. This logic covers targets, weapons, defense penetration, and damage assessment.
An algorithm does not invent this doctrine. It formalizes the choices and assumptions of its developers. An error in these assumptions can therefore be reproduced at high speed and scale.
The System Unlocks the Full Value of the PLAAF Fleet
Images broadcast by CCTV featured several major Chinese aircraft. The J-20, J-16, J-10C, and H-6K bomber were shown in the sequence alongside various missile systems.
These platforms fulfill complementary roles. The J-20 provides low observability, advanced sensors, and long-range combat capabilities. The J-16 can carry a heavy payload and conduct strike, escort, or electronic warfare missions depending on the variant. The J-10C offers a lighter multirole option. The H-6K allows for the deployment of long-range cruise missiles.
Artificial intelligence must unite these capabilities into systems warfare. The performance of an isolated aircraft becomes less important than its contribution to the whole.
A stealth fighter can detect a target without attacking it directly. It can transmit the position to another aircraft or to a missile launched from a standoff position. An airborne early warning aircraft can fuse data from multiple radars and redistribute the tactical picture to formations.
The documentary illustrates precisely an airborne detection and command aircraft directing a complex engagement. The scenario involved real targets, decoys, low-altitude flights, and intermittent electromagnetic interference. The aircraft reportedly identified enemy intent before guiding fighters toward a command node.
This sequence confirms the central role of airborne command platforms. They are no longer merely flying radars; they are becoming centers for information fusion, coordination, and redistribution.
MUM-T Links the J-20S to Combat Drones
The planning system fits into a broader shift toward MUM-T, or Manned-Unmanned Teaming. This concept refers to cooperation between piloted aircraft and uncrewed platforms.
In this model, a crewed aircraft does not necessarily micromanage every movement of a drone. It assigns a mission, receives data, and modifies priorities. The drone then executes certain tasks with varying degrees of autonomy.
China has officially presented the J-20S as an aircraft capable of participating in this type of operation. This twin-seat variant of the J-20 can accommodate a second crew member dedicated to managing sensors, data links, and uncrewed platforms.
Having two crew members is not a mere luxury. In high-intensity combat, a pilot must manage flight operations, threats, and weapon employment. Requiring the pilot to supervise multiple drones would severely increase cognitive load. The second crew member can focus on collaborative tactics and the overall tactical picture.
The GJ-11 Extends Range and Reduces Human Exposure
The GJ-11 is one of the platforms most frequently associated with Chinese ambitions in crewed-uncrewed teaming. This tailless aircraft adopts a flying-wing configuration designed to minimize radar signature.
In November 2025, the PLAAF released video footage showing an aircraft identified as the GJ-11 flying alongside a J-20 and a J-16D electronic warfare aircraft. This marked the first official depiction of such a formation.
In an operational setting, a drone of this type could fly ahead of the main formation. It could hunt for radars, gather intelligence, provoke defense activations, or strike high-threat installations.
The J-16D could jam radars and communications. The J-20 could hold back to fuse information, protect the formation, or engage aerial threats. The GJ-11 would absorb part of the risk that commanders prefer not to impose on human crews.
This pairing also multiplies the number of sensors. Multiple dispersed platforms observe the same area from different angles. They can compare radar emissions, images, and trajectories to improve target identification.
Public Demonstration Does Not Yet Prove Complete Mastery
A filmed formation does not automatically prove full operational capability. It does not demonstrate that the J-20 directly controlled the drone, that the aircraft exchanged all data seamlessly, or that the GJ-11 could adapt its mission autonomously.
Public research on Chinese capabilities highlights a lack of solid evidence regarding large-scale MUM-T deployment. Some exercises have shown drones transmitting coordinates to a ground base, which then passed them to fighters. This setup remains distinct from direct, dynamic control from the cockpit.
However, the system unveiled in August 2026 could accelerate this integration by providing a common planning layer. Drones, aircraft, and missiles can receive synchronized tasks even if tactical control remains distributed among different operators.
Multi-Domain Logic Extends Beyond Airpower
China is preparing operations where aerial, space, naval, land, cyber, and electromagnetic effects are combined. The system presented by CCTV is primarily a PLAAF tool; it is not yet publicly described as a complete joint force platform.
Nonetheless, its architecture aligns with the requirements of multi-domain operations. A modern airstrike relies on intelligence produced by satellites, drones, land radars, ships, and electronic signals intelligence stations.
A maritime target might initially be detected by a satellite. A drone can refine its coordinates. An airborne early warning aircraft can establish a higher-quality track. A bomber or ground battery can then launch the strike missile.
Artificial intelligence must select the optimal sensor and effector, regardless of whether both belong to the same military branch. This is the core principle of a distributed kill web.
Such an architecture would be critical in a high-intensity conflict in the South China Sea or around Taiwan. Distances, target numbers, and defense density would demand seamless coordination among the PLAAF, the People’s Liberation Army Navy, and the People’s Liberation Army Rocket Force.
The August 2026 presentation does not prove this joint integration is complete, but it shows that the PLAAF now possesses a digital building block designed to handle larger and faster operations.
Data Quality Is the Critical Single Point of Failure
The system’s effectiveness relies on reliable data. Artificial intelligence does not observe the battlefield directly; it processes information supplied by sensors and formatted according to predefined standards.
An incorrect radar position can lead to routing aircraft into contested airspace. An improperly identified target can result in wasted munitions. An erroneous damage assessment can trigger an unnecessary follow-up strike or leave a vital threat intact.
The issue worsens when data arrives asynchronously. A satellite image might be dozens of minutes old. A radar may temporarily lose a track. A drone might transmit a newer position, but with insufficient precision.
The system must therefore evaluate the freshness, reliability, and consistency of incoming data, distinguishing a genuine tactical shift from a measurement error.
Adversaries will specifically target this dependency. They may employ decoys, false radar emissions, mockups, rapid repositioning, or simulated traffic, making a high-value target appear where it is not.
A high-speed algorithm fed bad data does not produce a better decision; it produces a bad decision faster.
Cyberattacks Could Turn AI Against Its User
Cyber vulnerability represents another major weakness. The planning system becomes a primary target as soon as it centralizes information regarding forces, routes, and objectives.
An intrusion could seek to exfiltrate operational plans or subtly manipulate parameters. Altering a coordinate, delaying an order, or spoofing the status of a runway could suffice to throw an attack wave into disarray.
Data poisoning presents an even more insidious risk. An adversary injecting false information into software feeds could cause the system to allocate resources to fictitious targets or draw erroneous conclusions about enemy posture.
Machine learning models are also susceptible to adversarial inputs. Subtle modifications to an image or signal can deceive classification algorithms, causing an object to be ignored, misidentified, or classified as a higher threat than it is.
Finally, communication networks remain a critical vulnerability. MUM-T formations depend on data links, satellites, and airborne relays. High-power jamming can reduce bandwidth, delay information, or isolate a platform.
Chinese researchers themselves acknowledge these challenges. Military publications highlight the necessity of safeguarding algorithms, stress-testing models through continuous adversary simulations, and feeding exercise data back into the system to correct flaws.
Centralization Accelerates Strikes but Concentrates Risk
A unified system simplifies planning, but it can create overreliance on a limited number of computing and command nodes.
The more centralized the architecture, the greater the impact of a strike against a key node. Destroying an airborne early warning aircraft, jamming a satellite, or attacking a command center could degrade an entire campaign.
China must therefore build in redundancy. It must be possible to recalculate plans from multiple sites, units must remain able to operate with degraded communications, and drones must retain basic functionality if primary links fail.
This creates a tension between two models. Centralization allows efficient resource allocation across a campaign; decentralization improves survival and grants more initiative to frontline units.
Chinese command culture remains traditionally centralized. Artificial intelligence could reinforce this trend by providing higher echelons with a more detailed picture and granular control. Conversely, it could produce the opposite effect if tools transmit broad commander’s intent rather than detailed orders down to units.
This choice will not be purely technical; it will depend on the level of trust placed in subordinate commanders, aircrews, and autonomous systems.
The Decisive Test Will Come in a Truly Contested Environment
The system revealed in August 2026 represents a credible evolution in Chinese airpower. It addresses a concrete challenge: a staff can no longer manually plan, within tight timeframes, an operation involving hundreds of targets and over 100 units.
Artificial intelligence can accelerate information processing, weapon allocation, and wave coordination. It can integrate disparate platforms into a more cohesive force and prepare for the deployment of growing numbers of drones alongside piloted aircraft.
Beijing is thus pursuing decision superiority. Its goal is not simply to field more fighters or missiles, but to observe, decide, and strike before the adversary can reorganize its defenses.
However, the public demonstration does not allow for a measurement of the true operational maturity of this capability. Successful planning during a scripted exercise does not guarantee the same outcome when sensors are spoofed, networks jammed, and command centers targeted.
True progress will not be measured by the speed with which the software generates an initial plan. It will depend on its ability to detect dubious information, correct errors, and continue operating when the network begins to fragment.
Artificial intelligence can shorten the kill chain, but it also expands the digital attack surface of the force utilizing it. China has shown that it understands the first reality; it has yet to prove that it can master the second.
Live a unique fighter jet experience