U.S. Tests AI Integration on F-35 Fighters to Enhance Targeting and Threat Detection
{loadposition bannertop}
{loadposition sidebarpub}
Lockheed Martin flight tested an AI-enabled combat identification function on the F-35 at Nellis Air Force Base, demonstrating rapid onboard threat classification and same-cycle software updates under Project Overwatch. The test signals a push to shorten emitter recognition timelines and strengthen U.S. decision advantage in contested spectrum warfare.
Lockheed Martin has flight-tested an AI-enabled combat identification function on the F-35, aiming to compress the time it takes pilots to classify threats and act inside contested airspace. The demonstration, conducted under “Project Overwatch,” is significant not because the F-35 lacked sensor fusion, but because it shows a tactical AI model generating an independent identification output directly on the pilot’s display while airborne, then being rapidly updated on the ground in time for the next sortie. That update tempo is the real operational story: in a spectrum environment where radar modes, datalinks, and deception tactics change faster than traditional reprogramming cycles, the side that adapts first often survives.Follow Army Recognition on Google News at this link
Project Overwatch equips the F-35 with onboard AI that rapidly classifies unknown emitters, shortens threat identification time, enhances sensor fusion, and speeds mission data updates, strengthening survivability and decision advantage in contested airspace (Picture source: U.S. DoW).
Project Overwatch was flown at Nellis Air Force Base, Nevada, where a Lockheed-built and trained AI and machine learning model resolved identification ambiguities among emitters, improving situational awareness and reducing pilot decision latency. In practical terms, emitters point to radio-frequency sources such as surveillance radars, fire-control radars, and communications nodes, the very signatures that define modern integrated air defense systems and airborne intercept chains. The workflow is particularly noteworthy: engineers used an automated tool to label new emitters, retrained the model to recognize a new emitter class within minutes, and reloaded the updated model for the next flight within the same mission planning cycle. That is a fast-loop MLOps approach being pushed down to the tactical edge.
To understand why this matters, it helps to anchor the AI in the F-35’s architecture. The aircraft’s combat power is built around fusing diverse sensors into a single tactical picture and then presenting it in a way a pilot can exploit at speed. Core contributors include the AN/APG-81 AESA radar, capable of air-to-air and air-to-ground modes while supporting electronic protection and electronic attack functions; the AN/ASQ-239 Barracuda electronic warfare suite providing 360-degree threat awareness and countermeasures; the Distributed Aperture System delivering spherical infrared coverage; and the internally mounted Electro-Optical Targeting System for precision targeting and long-range detection. The human-machine interface, including the helmet display system, is designed to turn that fused data into actionable cues rather than raw sensor clutter. Overwatch effectively adds an adaptive classifier into the same decision pipeline that already converts RF and IR observations into icons, priorities, and engagement recommendations.
The near-term advantage is speed and confidence in threat characterization, especially when the RF environment is ambiguous. A novel or modified emission can signal a new system, a known system in a new mode, or a decoy designed to pull shooters out of position. Today, the U.S. Air Force runs a global process in which collected emissions are analyzed and turned into updated mission data files for aircraft, a cycle that has improved dramatically but still depends on collection, processing, verification, distribution, and cockpit integration. Project Overwatch represents an attempt to accelerate and partially automate that chain, reducing the gap between discovery of a new signal and its operational exploitation. Notably, the capability was developed using internal research and development funding, signaling an industrial push to shape the next phase of F-35 software evolution.
This configuration can tighten the kill chain in at least three ways. First, it reduces pilot workload by collapsing “what is it” time, which is increasingly the gating factor when engagements occur beyond visual range and under emissions control. Second, it can improve survivability by allowing stealth aircraft to remain passive longer. Faster passive identification and prioritization mean fewer confirmatory emissions and fewer seconds exposed to counter-detection. Third, it can enhance coordinated fires. F-35s are optimized to share tracks and target-quality information across formations and to other platforms; their low probability of intercept networking concepts are intended to preserve survivability while still distributing the tactical picture. An AI-derived classification attached to an emitter track is precisely the type of data that can cue jamming, standoff weapons, or surface-based fires, provided it meets confidence thresholds and rules of engagement. This is where AI among the fighters becomes meaningful: not autonomous weapons release, but distributed, faster-than-human sensemaking propagated through the force.
Development context also matters: the F-35 is in the middle of a compute-driven modernization arc in which software ambition is constrained or enabled by onboard processing power. Technology Refresh 3 is designed to increase processing capacity and memory headroom to support Block 4 upgrades, including stronger data fusion and enhanced electronic warfare capabilities. While the TR-3 schedule has faced delays, the strategic direction remains clear: the F-35’s next leaps are software-defined, and that increasingly includes AI-enabled functions that must run on the aircraft itself. The fact that Overwatch’s model is compact enough to operate onboard underscores its practical orientation toward deployable capability rather than conceptual demonstration.
Zooming out, Project Overwatch sits within a broader U.S. effort to operationalize artificial intelligence across the joint force. The Department of Defense has framed data, analytics, and AI as central to achieving decision advantage in high-end conflict, with sustained budget lines dedicated to AI development and integration. The Pentagon has expanded its governance structures and acquisition pathways to accelerate fielding while attempting to manage risk, accountability, and verification. For combat aircraft such as the F-35, this means AI will not remain a discrete pilot program. It will be embedded into mission systems in the same way that sensor fusion defined the fifth-generation paradigm.
There are, however, hard questions before Overwatch becomes an operational baseline. Combat identification is a high-consequence function in which false positives can increase fratricide risk, trigger escalation, or waste scarce munitions. An adaptive model retrained at high speed must still be testable, auditable, and resilient against adversary deception, particularly in the electronic warfare domain where spoofing and signature manipulation are core tactics. The most credible path forward is incremental: AI as a confidence-graded assistant that accelerates analyst-to-cockpit updates and highlights ambiguity rather than replacing human judgment.
If that balance is achieved, the payoff is considerable. With more than a thousand F-35s fielded globally across U.S. and allied forces, even a modest reduction in identification latency and reprogramming timelines translates into a measurable advantage in the opening phase of a high-intensity air campaign. In a fight defined by contested spectrum, compressed timelines, and adaptive adversaries, Project Overwatch signals that the United States is not only building advanced fighters but embedding adaptive intelligence directly into the heart of its air combat architecture.
Written by Evan Lerouvillois, Defense Analyst.
Evan studied International Relations, and quickly specialized in defense and security. He is particularly interested in the influence of the defense sector on global geopolitics, and analyzes how technological innovations in defense, arms export contracts, and military strategies influence the international geopolitical scene.
{loadposition bannertop}
{loadposition sidebarpub}
Lockheed Martin flight tested an AI-enabled combat identification function on the F-35 at Nellis Air Force Base, demonstrating rapid onboard threat classification and same-cycle software updates under Project Overwatch. The test signals a push to shorten emitter recognition timelines and strengthen U.S. decision advantage in contested spectrum warfare.
Lockheed Martin has flight-tested an AI-enabled combat identification function on the F-35, aiming to compress the time it takes pilots to classify threats and act inside contested airspace. The demonstration, conducted under “Project Overwatch,” is significant not because the F-35 lacked sensor fusion, but because it shows a tactical AI model generating an independent identification output directly on the pilot’s display while airborne, then being rapidly updated on the ground in time for the next sortie. That update tempo is the real operational story: in a spectrum environment where radar modes, datalinks, and deception tactics change faster than traditional reprogramming cycles, the side that adapts first often survives.
Follow Army Recognition on Google News at this link
Project Overwatch equips the F-35 with onboard AI that rapidly classifies unknown emitters, shortens threat identification time, enhances sensor fusion, and speeds mission data updates, strengthening survivability and decision advantage in contested airspace (Picture source: U.S. DoW).
Project Overwatch was flown at Nellis Air Force Base, Nevada, where a Lockheed-built and trained AI and machine learning model resolved identification ambiguities among emitters, improving situational awareness and reducing pilot decision latency. In practical terms, emitters point to radio-frequency sources such as surveillance radars, fire-control radars, and communications nodes, the very signatures that define modern integrated air defense systems and airborne intercept chains. The workflow is particularly noteworthy: engineers used an automated tool to label new emitters, retrained the model to recognize a new emitter class within minutes, and reloaded the updated model for the next flight within the same mission planning cycle. That is a fast-loop MLOps approach being pushed down to the tactical edge.
To understand why this matters, it helps to anchor the AI in the F-35’s architecture. The aircraft’s combat power is built around fusing diverse sensors into a single tactical picture and then presenting it in a way a pilot can exploit at speed. Core contributors include the AN/APG-81 AESA radar, capable of air-to-air and air-to-ground modes while supporting electronic protection and electronic attack functions; the AN/ASQ-239 Barracuda electronic warfare suite providing 360-degree threat awareness and countermeasures; the Distributed Aperture System delivering spherical infrared coverage; and the internally mounted Electro-Optical Targeting System for precision targeting and long-range detection. The human-machine interface, including the helmet display system, is designed to turn that fused data into actionable cues rather than raw sensor clutter. Overwatch effectively adds an adaptive classifier into the same decision pipeline that already converts RF and IR observations into icons, priorities, and engagement recommendations.
The near-term advantage is speed and confidence in threat characterization, especially when the RF environment is ambiguous. A novel or modified emission can signal a new system, a known system in a new mode, or a decoy designed to pull shooters out of position. Today, the U.S. Air Force runs a global process in which collected emissions are analyzed and turned into updated mission data files for aircraft, a cycle that has improved dramatically but still depends on collection, processing, verification, distribution, and cockpit integration. Project Overwatch represents an attempt to accelerate and partially automate that chain, reducing the gap between discovery of a new signal and its operational exploitation. Notably, the capability was developed using internal research and development funding, signaling an industrial push to shape the next phase of F-35 software evolution.
This configuration can tighten the kill chain in at least three ways. First, it reduces pilot workload by collapsing “what is it” time, which is increasingly the gating factor when engagements occur beyond visual range and under emissions control. Second, it can improve survivability by allowing stealth aircraft to remain passive longer. Faster passive identification and prioritization mean fewer confirmatory emissions and fewer seconds exposed to counter-detection. Third, it can enhance coordinated fires. F-35s are optimized to share tracks and target-quality information across formations and to other platforms; their low probability of intercept networking concepts are intended to preserve survivability while still distributing the tactical picture. An AI-derived classification attached to an emitter track is precisely the type of data that can cue jamming, standoff weapons, or surface-based fires, provided it meets confidence thresholds and rules of engagement. This is where AI among the fighters becomes meaningful: not autonomous weapons release, but distributed, faster-than-human sensemaking propagated through the force.
Development context also matters: the F-35 is in the middle of a compute-driven modernization arc in which software ambition is constrained or enabled by onboard processing power. Technology Refresh 3 is designed to increase processing capacity and memory headroom to support Block 4 upgrades, including stronger data fusion and enhanced electronic warfare capabilities. While the TR-3 schedule has faced delays, the strategic direction remains clear: the F-35’s next leaps are software-defined, and that increasingly includes AI-enabled functions that must run on the aircraft itself. The fact that Overwatch’s model is compact enough to operate onboard underscores its practical orientation toward deployable capability rather than conceptual demonstration.
Zooming out, Project Overwatch sits within a broader U.S. effort to operationalize artificial intelligence across the joint force. The Department of Defense has framed data, analytics, and AI as central to achieving decision advantage in high-end conflict, with sustained budget lines dedicated to AI development and integration. The Pentagon has expanded its governance structures and acquisition pathways to accelerate fielding while attempting to manage risk, accountability, and verification. For combat aircraft such as the F-35, this means AI will not remain a discrete pilot program. It will be embedded into mission systems in the same way that sensor fusion defined the fifth-generation paradigm.
There are, however, hard questions before Overwatch becomes an operational baseline. Combat identification is a high-consequence function in which false positives can increase fratricide risk, trigger escalation, or waste scarce munitions. An adaptive model retrained at high speed must still be testable, auditable, and resilient against adversary deception, particularly in the electronic warfare domain where spoofing and signature manipulation are core tactics. The most credible path forward is incremental: AI as a confidence-graded assistant that accelerates analyst-to-cockpit updates and highlights ambiguity rather than replacing human judgment.
If that balance is achieved, the payoff is considerable. With more than a thousand F-35s fielded globally across U.S. and allied forces, even a modest reduction in identification latency and reprogramming timelines translates into a measurable advantage in the opening phase of a high-intensity air campaign. In a fight defined by contested spectrum, compressed timelines, and adaptive adversaries, Project Overwatch signals that the United States is not only building advanced fighters but embedding adaptive intelligence directly into the heart of its air combat architecture.
Written by Evan Lerouvillois, Defense Analyst.
Evan studied International Relations, and quickly specialized in defense and security. He is particularly interested in the influence of the defense sector on global geopolitics, and analyzes how technological innovations in defense, arms export contracts, and military strategies influence the international geopolitical scene.
