13 January 2020, early morning before dawn; the account estimates about two minutes
Outcome: Open (Unresolved). GEIPAN retained D after fieldwork, expert review and enquiries at Lanvéoc-Poulmic. No satisfactory aircraft, drone or lighthouse identification was established; reporting and reconstruction delays remain explicit limitations.
Three rectangles in bad weather
A driver on the Crozon peninsula described seeing two bright red horizontal rectangles beside the road, then a third appearing at the same spacing. He reported stopping near a junction while the lights remained in view, before they moved away towards the south. The account comes from 13January 2020, during strong wind and rain.
The witness did not describe a visible body joining the lights. Three features maintaining their spacing do not, by themselves, establish a triangular craft or one rigid vehicle. The catalogue therefore uses Multiple lights without adding a triangle or disc tag.
A long investigation of a short event
GEIPAN received the questionnaire inJune 2020 and visited the location inJanuary 2021. Additional questions, expert discussion and exchanges with the nearby naval air station followed. The detailed report is datedJuly 2023. Its length documents investigative work; it is not the length of an event recording.
The report also states why those delays matter. Descriptions can change between an event, a questionnaire written months later and a reconstruction more than a year afterwards. A careful account should retain that qualification alongside the agency’s favourable assessment of the witness’s cooperation.
Aircraft, drones and coastal lights
Investigators examined helicopters, drones, aircraft, lighthouse or semaphore lighting, wind-turbine lights, projected beams, vehicle lights and reflections inside the van. Enquiries at Lanvéoc-Poulmic did not establish a relevant flight or drone exercise by that base.
That negative response narrows one source of possible activity. It does not exclude aircraft from every other operator. The report also tested how sound, apparent dimensions, weather and the reported movement fitted the more plausible alternatives, rather than relying solely on a witness’s conviction that the appearance was unusual.
The sound comparison needs its surroundings
The report considered the vehicle’s running engine, closed windows, rain, wind direction and the sound expected from a nearby helicopter. It judged the combination difficult for that explanation. This is a reasoned comparison with the account, not an acoustic recording or a guarantee that every aircraft at every distance would have been heard.
The same caution applies to the reported sudden departure. Without an independently established range, a rapid change in angular appearance cannot be converted into an exact acceleration. The careful result is that the sequence remained unexplained, not that an extraordinary flight performance was measured.
A reconstruction is not original footage
GEIPAN produced a visual simulation to communicate the reported sequence. It should be viewed as a reconstruction, not as a film made during the sighting. The same distinction applies to later annotated setting photographs: they can locate sightlines while supplying no independent image of the lights.
Following remote enquiry, fieldwork and expert review, GEIPAN retained D because no considered explanation adequately fitted the account. UKPX records Open (Unresolved). The surviving uncertainty concerns both identity and how accurately the remembered appearance maps onto physical sources. Better contemporary imagery or a documented matching operation could alter the result.
Location and map reference
Crozon, Finistère, France. Broad commune reference from the French government geographic API; not the witness home, precise observation position or object location. This listing does not advertise access to private property.
Sources and review
CNES–GEIPAN official case record and the original investigation report. Official summary;70-page report pp 1–6 and 68–70 reviewed for narrative, chronology, alternatives and cautions. No claim all 70 pages/attachments or simulation video forensically analysed. Sources and current classification checked 13 November 2026. The incident date above is distinct from the later investigation and archive-update dates.