AI fighter jets are moving from science-fiction imagery to instrumented flight tests in 2026, but the decisive change is not the disappearance of human pilots. It is the emergence of a combat-air model in which a crewed fighter can coordinate one or more uncrewed aircraft, while autonomy software handles selected flight and mission tasks.
The United States has tested an F-22 directing a collaborative combat aircraft and an Anduril YFQ-44A firing an AIM-120 missile. Türkiye has flown Bayraktar KIZILELMA in autonomous formations with Leonardo’s M-346, while Airbus and Kratos completed the U740 Valkyrie’s first European flight. These are concrete development milestones — not proof that independently fighting AI squadrons have entered service.
What Is an AI Fighter Jet? Three Technologies Are Often Confused
The term “AI fighter jet” covers at least three distinct technologies. The first is a human-piloted aircraft with AI-supported sensor fusion, decision aids or mission planning. The second is a conventional jet adapted to let an AI agent control its flight, as demonstrated in research programmes such as the X-62A VISTA and 2026 VENOM F-16 experiments. The third is an uncrewed combat aircraft designed from the outset to coordinate with crewed aircraft — usually called a Collaborative Combat Aircraft (CCA).
They should not be described interchangeably. An AI-controlled demonstration aircraft is not necessarily an operational combat platform; an unmanned fighter does not necessarily select targets independently; and a sixth-generation crewed fighter is not automatically an autonomous aircraft.
U.S. CCA Programme: YFQ-42A and YFQ-44A Move Beyond Basic Flight
The U.S. Air Force is developing the General Atomics YFQ-42A and Anduril YFQ-44A Fury as candidate collaborative combat aircraft. In October 2026, an F-22 Raptor pilot directed a YFQ-44A during an airborne tactical-control demonstration, showing how a crewed fighter can assign mission tasks without remotely piloting every manoeuvre.
On 15 July, the Air Force reported that the YFQ-44A fired an AIM-120 missile at a digital target. The test advanced weapons integration, but it was not the destruction of a real aircraft and did not establish authorization for unsupervised lethal decisions. The programme retained human oversight of weapon release.
The YFQ-42A programme also illustrates how difficult trustworthy autonomy remains. General Atomics said testing resumed in May 2026 after an April aircraft loss attributed to an autopilot weight-and-centre-of-gravity calculation problem. That incident is a caution against interpreting prototype demonstrations as evidence of a fully mature operational system.
Why AI Software May Matter More Than the Airframe
The Air Force is using its Autonomy Government Reference Architecture (A-GRA) to separate mission-autonomy software from aircraft hardware and allow multiple vendors to integrate capabilities. This creates a potential route for rapid upgrades and competition across sensors, autonomy algorithms and aircraft designs rather than locking one software stack to one platform.
It also exposes a key vulnerability. Future CCAs must keep useful autonomy when satellite navigation, data links and communications are degraded, and their commanders must understand what the system will do if a connection disappears. Open architecture is a procurement and integration strategy, not a guarantee of survivability.
AI-Piloted F-16s: From X-62A VISTA to VENOM
DARPA’s Air Combat Evolution programme put AI agents in control of the modified X-62A VISTA during fighter-manoeuvre trials against a human-piloted F-16, establishing an important research baseline. In July 2026, DARPA said VENOM-modified F-16s were undergoing AI-controlled flight testing, with human safety pilots remaining in the cockpit to monitor and retake control.
The distinction between the aircraft moving under AI control and a weapon being authorized to engage a target is fundamental. Experiments can validate manoeuvres, model robustness and human-machine cooperation without transferring legal or operational responsibility to the software.
Türkiye’s KIZILELMA Tests Autonomous Wingman Operations
Türkiye’s Bayraktar KIZILELMA is one of the most visible uncrewed fighter development programmes. In a joint June 2026 statement, Baykar and Leonardo described live K-SWARM trials in which an M-346 pilot commanded formations and manoeuvres that KIZILELMA executed autonomously.
Baykar has also reported a November 2025 GÖKDOĞAN air-to-air missile test against a target aircraft, and July 2026 tests of TEBER-82 and TOLUN weapons carried inside KIZILELMA’s weapons bay. Those trials demonstrate distinct elements of weapon integration and autonomous collaboration; they do not by themselves establish autonomous target selection or routine operational squadron deployment.
K-SWARM has wider industrial significance: a Turkish unmanned combat aircraft and an Italian crewed aircraft worked through shared data exchange, command interfaces and autonomy algorithms. Interoperability — especially among aircraft made by different manufacturers — is emerging as a central strategic requirement for next-generation fleets.
Airbus U740 Valkyrie Shows Europe’s Near-Term Approach
On 12 October 2026, the Airbus–Kratos U740 Valkyrie completed its first European flight. The aircraft used a runway-independent rocket-assisted launch and flew at speeds up to Mach 0.8. The company is working toward a combat-ready European CCA by 2029.
The core European contribution is intended to be Airbus’s MARS (Multi-platform Autonomous Reconfigurable and Secure) mission system. It was not installed for the first European flight, according to Reuters, and more demanding MARS-enabled testing is planned. The 2029 date is an industrial objective rather than an achieved operational milestone.
Sixth-Generation Fighters: F-47, F/A-XX and GCAP Are Different Programmes
In parallel with uncrewed wingmen, governments are developing new crewed fighters. Boeing won the U.S. Air Force F-47 programme in 2025, and the aircraft is still in development with a reported first-flight target of 2028. In September 2026, the U.S. Navy selected Boeing for its separate F/A-XX sixth-generation fighter programme. Neither programme should be described as a fielded AI-piloted jet.
The UK, Italy and Japan are working on the Global Combat Air Programme (GCAP), which targets service entry in 2035. A £4.6 billion design contract announced in July 2026 is a major industrial commitment, and the UK Ministry of Defence says AI and digital engineering are part of the work. The final autonomy and engagement architecture of the eventual aircraft has not been publicly demonstrated.
Can AI Fighter Jets Operate Without a Pilot or Data Link?
Autonomous aircraft can perform some navigation, formation control and mission changes with limited manual input. Yet combat requires much more than stable flight: the aircraft must maintain a usable tactical picture, identify friend from foe, respond to electronic attack, conserve fuel and munitions, and behave predictably when its data link is interrupted.
Another design tension is cost. An aircraft advertised as affordable mass still needs propulsion, sensors, communications, maintenance support and mission assurance. For CCAs, the relevant metric is not just purchase price, but how much operational effect the force can generate per aircraft lost, sortie flown and aircraft maintained.
Human Control and the Law Remain Central
The spread of AI-assisted flight must not be confused with a legal green light for autonomous targeting. The International Committee of the Red Cross distinguishes autonomous weapon systems by their ability to select and apply force without further human intervention and argues for additional international restrictions. Existing international humanitarian law still governs attacks regardless of the technology used.
Autonomous manoeuvring, target recognition, a pilot’s approval of a weapon release and a weapon system that independently chooses its own targets represent different levels of control. Public reports of a “dogfight win” or a missile test should not be read as evidence that a programme has resolved those legal, operational and ethical questions.
What to Watch From Late 2026 to 2035
The near-term tests to watch are continued YFQ-42A and YFQ-44A flight and weapons integration, Airbus MARS-enabled Valkyrie trials, expansion of KIZILELMA’s K-SWARM work and higher-fidelity human-autonomy evaluation using F-16 testbeds. Longer-term milestones are F-47 flight testing, F/A-XX development and GCAP’s 2035 service target.
The central lesson of 2026 is that the future of air combat looks less like a single robot fighter replacing the pilot and more like a networked force of crewed jets, autonomous aircraft, software-defined mission systems and human command authority. The aircraft are progressing rapidly; the hard work of certification, battlefield reliability, trust, production and accountable control remains.
