AI controversy forces Nikon to replace its science video competition winner

Nikon has disqualified the winning entry in its 2026 Small World in Motion competition after determining that it breached the contest’s rules on generative artificial intelligence.

The decision means Nguyen Nam Nhat’s video, originally awarded second place, has been promoted to first.

The original winner, Dr Ning Xu, submitted footage depicting tiny hair-like structures called cilia inside the airways. The footage showed abnormal cilia movement associated with primary ciliary dyskinesia (PCD), a rare inherited condition in which defective cilia can make it difficult for the respiratory system to clear mucus and bacteria.

The entry was produced using diffractive super-resolution microscopy, a technique capable of revealing biological structures at a scale that conventional optical microscopy cannot resolve clearly. That made the footage scientifically significant as well as visually striking: viewers were seeing the movement of structures that are normally invisible.

Why Nikon disqualified the entry

The dispute centred on the processing applied to the original microscopy data.

Xu acknowledged using an unsupervised neural-network method during post-processing. He said the system was used to help distinguish and visualise features in the reconstructed images rather than to create the underlying cilia or their movement.

That distinction is important because scientific images routinely undergo processing. Researchers may remove noise, improve contrast, reconstruct information or use computational methods to make structures within raw data visible. Those processes do not automatically make an image artificial.

Generative AI creates a different problem because a system can potentially introduce or reconstruct visual information that was not directly recorded.

Nikon ultimately determined that the processing used in Xu’s entry crossed the boundary set by its competition rules. The company reviewed the submission and supporting material before withdrawing the prize.

The problem with AI and scientific images

The case exposes a difficult distinction for image competitions. A photograph or video can be heavily processed while still representing something that was genuinely captured. The problem arises when software begins determining what missing, unclear or ambiguous visual information should look like.

That matters particularly in scientific microscopy. A viewer may assume that every visible structure represents something observed through the microscope, when sophisticated computational processing can alter how that structure appears.

The finished image alone may therefore no longer provide enough information to establish what was actually captured.

For competition organisers, checking authenticity becomes harder as machine-learning techniques become more sophisticated. A blanket ban on all AI-assisted processing could exclude legitimate scientific imaging methods, while a narrow definition of generative AI could leave room for systems that substantially reconstruct what viewers see.

What happens next

Nikon has said it will review its competition rules and evaluation procedures as AI technology develops. That could lead to greater emphasis on original files and processing records in future competitions. Judges may need to establish not only whether AI was used, but exactly what the technology did to the underlying image.

The 2026 result therefore raises a question that extends beyond Nikon’s competition: when software is capable of reconstructing what a microscope cannot clearly capture, where should the line be drawn between enhancement and generation?

For scientific imaging competitions, that distinction is becoming increasingly important because the credibility of an entry depends not only on how extraordinary the final footage looks, but on whether viewers can trust that what they are seeing was actually present in the original observation.