Nikon reviews winning video after researchers allege AI anomalies

Generative artificial intelligence tools are facing intense scrutiny within scientific photography competitions after judges awarded a top microscopic video that researchers later argued displayed biologically impossible structures. The controversy centers on the 16th edition of the Nikon Small World in Motion competition, where a winning entry depicting respiratory cilia movement has triggered a formal review of its post-production methods by organizers and raised urgent questions about data integrity in scientific imaging.

In September, Nikon announced the winner of its contest from 347 total submissions spanning 40 countries. Dr Ning Xu, formerly associated with Tsinghua University and now working as a research fellow at the National University of Singapore, secured first place with a detailed video recording. His footage illustrates incorrect cilia movement in the airways of a child with primary ciliary dyskinesia (PCD). The recording relied on super-resolution microscopy alongside a digital micromirror device to capture high-speed cellular motion.

Edward Phelps and Ian Donovan Identify Anomalies

Days after the announcement, Edward Phelps of the University of Florida posted observations on LinkedIn regarding visual inconsistencies and a lack of image continuity throughout the clip. Phelps pointed out that distinct purple structures appeared and vanished intermittently, while other elements mirrored mitochondria the size of cell nuclei situated outside cells—an organization entirely absent from biology. Ian Donovan from UT Southwestern Medical Center added that running the video through Google Gemini reportedly yielded SynthID, an invisible watermarking system utilized by Google to identify AI-generated media. Neither Google nor Nikon has publicly verified that specific watermark reading.

Nikon updates winning entry description after AI use emerges

On September 22, Nikon updated the winning entry’s description to reflect that Xu employed an unsupervised AI model during post-production. Xu acknowledged using artificial intelligence but maintained that the technology did not generate the video, the cilia, or their physical motion. Instead, he explained that the model was applied to reconstructed grayscale data to differentiate structures possessing similar morphology and to assign colors. This specific extent of digital intervention quickly transformed into the core dispute of the competition.

Competition Organizers Launch Review of Submissions

Nikon stated on September 26 that it is actively reviewing the initial verification details alongside additional documentation provided by Xu. The competition organizers confirmed that the researcher is cooperating by supplying comprehensive information concerning his hardware, imaging protocols, and processing techniques. While contest rules prohibit AI-generated videos, judges must now determine the exact boundary of what the algorithmic model executed during post-processing. By October 1, Xu’s profile on LinkedIn was no longer accessible. Resolving the dispute would require comparing the award-winning video against raw grayscale data to isolate authentic microscopic captures from modifications introduced during processing, while Nikon has yet to officially declare a rule violation.

Frequently Asked Questions About the Nikon Small World Controversy

Which organization runs the Small World in Motion competition?

Nikon organizes the contest, which has shown the beauty of the world under a microscope for half a century.

What specific medical condition does the winning video attempt to show?

The winning entry depicts incorrect cilia movement inside the airways of a child with primary ciliary dyskinesia.

What tool did Ian Donovan use to check the disputed video file?

Ian Donovan stated that he ran the file through Google Gemini to check for SynthID, an invisible digital watermark utilized by Google for identifying AI-generated content.