What Dassault Isn’t Saying About Its New Rafale AI

Dassault says two homegrown AI algorithms have flown on the Rafale and are mature enough for future upgrades. It won’t say what they do or which jets get them first.

Dassault Aviation said Tuesday it has flight-tested two “sovereign” artificial intelligence algorithms aboard its Rafale fighter jet, one built by its own engineers and the other with Thales’ AI accelerator, cortAIx.

The French planemaker said both algorithms are ready to be considered for future versions of the jet. “They have reached a level of maturity that makes them eligible for future Rafale upgrades,” the company said in a statement.

Dassault did not say what the algorithms do or which Rafale standard will be the first to carry them.

Why the Dassault Rafale AI tests matter

The flight tests come as France prepares for its next generation of fighter jets. Those preparations follow the effective collapse in June of the Franco-German-Spanish Future Combat Air System, or FCAS, a joint program that involved Airbus.

Dassault described the new functions as part of a broader effort to bring controlled and supervised AI into the cockpit to serve the human crew.

The company also laid out why that is hard to do in a fighter.

“This capability requires mastering several key challenges specific to military aviation: ensuring the availability and quality of operational data (real or simulated), leveraging and synergizing domain expertise, and optimizing resource efficiency on an embedded platform subject to stringent constraints,” Dassault said.

Put simply, the software must work within the limited computing resources aboard the jet and rely on dependable operational data, whether gathered in real flights or in simulation.

From Thales partnership to flight test

The flight tests follow a strategic partnership Dassault Aviation and Thales signed Nov. 18, 2025. According to the Thales announcement, the companies agreed to develop controlled and supervised AI for crewed and uncrewed military aircraft, covering observation, situation analysis, decision-making, planning and control during military operations.

Thales is separately adding AI to the Rafale’s TALIOS targeting pod. Thales said in November 2024 that the feature would enter service with the Rafale F4.3 standard in 2026.

Thales has said the pod’s embedded AI analyzes images in real time aboard the aircraft and can search an area for objects of interest 100 times faster.

Harmattan AI and the drone connection

Dassault has also invested outside its traditional supplier base. On Jan. 12, it led a $200 million Series B funding round for French defense startup Harmattan AI, which valued the company at $1.4 billion, to develop embedded autonomy for the Rafale F5 and uncrewed combat air systems.

“Dassault Aviation has always placed technological excellence and sovereignty at the heart of its values. This partnership with Harmattan AI reflects our commitment to integrating high-value autonomy into the next generation of combat air systems,” Éric Trappier, Dassault Aviation’s chairman and CEO, said at the time.

Harmattan AI’s chief executive said the startup has no plans to build its own fighters.

“We don’t intend to build fighter jets. We are focusing on other types of attritable systems … but our autonomy stack actually has a lot of value for legacy systems that are still very relevant on the battlefield today, especially fighter jets and unmanned fighters in the future,” Mouad M’Ghari, CEO of Harmattan AI, told Tectonic Defense.

The two companies have already flown together. In a flight test announced July 13, a drone carrying their jointly developed NAMIB electronic warfare payload detected and located a radar, then passed its position to a Rafale F4, which simulated a strike on the target.

Rafale F5 moves to center stage after FCAS

The FCAS program, a roughly €100 billion effort to build a sixth-generation European fighter, was effectively terminated June 8. The partners could not agree on how to divide the work, who would own the intellectual property and what the aircraft should be able to do, disputes that centered on Dassault and Airbus.

Airbus Defence and Space had made no public statement about Dassault’s AI flight tests as of Tuesday.

France’s defense procurement agency, the Direction générale de l’armement, or DGA, announced Sept. 10 that it had awarded preliminary development contracts for the Rafale F5 standard.

The work covers a new Thales RBE2-XG active electronically scanned array radar with more power and detection range, better identification of targets with extremely low radar signatures and added computing capacity for AI. It also includes a fully digital SPECTRA electronic warfare suite from Thales and MBDA, and a broadband, jam-resistant data link to keep aircraft communicating in hostile airspace.

Safran Aircraft Engines received a contract for preliminary design of the M88 T-REX engine. The upgrade is meant to raise thrust 20%, from 7.5 metric tons to 9 metric tons, or from about 16,500 pounds to about 19,800 pounds.

The F5 is also set to carry MBDA’s ASN4G hypersonic nuclear cruise missile, replacing the upgraded ASMPA-R as the airborne leg of France’s nuclear deterrent.

The DGA aims to launch the F5’s full development phase by the end of 2026, with qualification targeted for around 2033 before the standard enters service with the French Air and Space Force and then the French navy.

How other militaries are testing AI in fighters

Dassault is not alone. Under the VENOM program, the U.S. Air Force and the Defense Advanced Research Projects Agency, or DARPA, are flight-testing modified F-16s in which an AI agent flies the jet while a human pilot on board monitors it.

In Sweden, Saab and German startup Helsing flight-tested an AI agent called Centaur aboard a Gripen E between May 28 and June 3, 2025. Saab said the software maneuvered the fighter on its own during simulated beyond-visual-range combat.

Unlike Saab, Dassault has not described what its algorithms did in the air.