Rafale AI flight tests reach the upgrade conversation

Dassault Aviation says it has flight-tested two sovereign artificial-intelligence algorithms on a Rafale fighter jet. One was developed by Dassault engineers and the other with Thales's cortAIx unit. The manufacturer describes both as mature enough to be considered for future Rafale upgrades, moving the work from a partnership announcement into an airborne test environment.

That wording is deliberately narrower than a deployment announcement. Dassault did not identify the algorithms' tasks, the aircraft standard used for the tests, the number of flights or a programme date. Reuters independently reported the tests and the absence of functional detail. Eligibility for a future upgrade means the technology can enter further evaluation; it does not show that it has been ordered, fielded or authorised for operational use.

What the announcement does not establish

The company says the functions are intended to support the human crew through controlled and supervised AI. Its release points to the distinctive engineering problems of military aviation: reliable real or simulated operational data, combining domain expertise and running computation on an embedded platform with strict resource limits. Those statements explain the design claim, but they do not disclose how human supervision was tested.

AeroTime connects the flights to a strategic partnership signed by Dassault and Thales in November 2025. It also notes the wider direction of Rafale development toward collaborative combat, in which crewed aircraft may work with uncrewed systems. The new tests can be read as one building block in that direction, not evidence that an autonomous wingman capability was demonstrated in these flights.

Integration is more than an algorithm

The competitive context is moving quickly. Reuters cites earlier demonstrations involving Saab's Gripen E and a modified US F-16 to show that manufacturers are testing AI assistance and machine control in real aircraft. Comparisons remain limited because companies disclose different tasks, safety arrangements and test conditions. A flight test proves integration under the conditions used; it does not create a common measure of operational capability.

Thales already supplies major Rafale systems and presents AI-based assistance as part of its air-combat portfolio. That existing industrial relationship matters because bringing software into an aircraft is also a certification, cybersecurity, data-governance and lifecycle-support problem. The value of a sovereign algorithm depends not only on who wrote it, but on how updates, training data and operational responsibility are controlled over time.

The next useful disclosures would identify the functions tested, the human decision boundary, safety evaluation and the Rafale standard that could receive them. For now, the verified milestone is significant but bounded: two French-developed algorithms flew on the aircraft, Dassault sees sufficient maturity for upgrade consideration, and no operational adoption has been announced.

Written for WHIF from the linked material. This article does not claim on-the-ground reporting. Our editorial standards.