In a dramatic turn for Nikon’s five‑year‑old Small World In Motion photo contest, the short film that took top honours is now being re‑examined after a group of scientists raised serious concerns that it was manipulated with artificial‑intelligence (AI) tools.
The winning entry, submitted by Dr Ning Xu of Tsinghua University, shows what he claims to be cilia—tiny, hair‑like organelles—moving inside the lungs of a child with a rare disease. The video was brightened with the colour palette that is allowed in the competition, but the form and motion of the structures have spurred debate.
While Dr Xu has denied that generative AI produced the cilia or their movement, he conceded that AI was applied to enhance the grayscale stack to a vivid, “high‑contrast” image. He insists that the raw data remained unchanged and that the contest rules were not breached.
Experts from Florida University, UT Southwest Medical Centre and former judges have found a “distinctive watermark” and a range of visual features that, according to them, “pop in and out” of existence, suggesting algorithmic alterations. Edward Phelps, an associate professor at the University of Florida, argued that the cilia appeared to emerge out of nowhere, while PhD student Ian Donovan points to the watermark as evidence of AI generation.
Nikon has stated that the entrant is “respectfully complying” with the investigation and has provided additional technical documents. In a statement, the company said it has found no rule violations to date and will issue an update once its review is complete.
The controversy presents a broader dilemma for scientific and artistic photography. Dr Patrick Hickey, a former winner of the contest, warned that the authenticity crisis could undermine the credibility of future submissions, especially in an era where AI can easily repair or enhance microscopic imagery.
As the investigation unfolds, the scientific community is closely watching whether Nikon will uphold its credibility and the standards that the Small World contest has long set.
