Nikon Strips Prize From A.I.-Altered Microscopy Video

Nikon has stripped the winner of its Small World in Motion microscopy contest after deciding the entry broke its rules on generative A.I. The company said the decision was not meant as a judgment on the entrant’s professional standing, his scientific work, or his intent. But the message was clear: the video crossed a line Nikon says the competition does not allow.

The now-disqualified winner was Ning Xu, an optical engineer at the National University of Singapore. His video showed cilia in airway tissue from a child with primary ciliary dyskinesia, a rare genetic disorder. Once it hit LinkedIn, scientists began picking it apart. The purple structures in the clip raised immediate suspicion, because they looked biological without actually matching anything known in biology.

Edward Phelps, a bioengineering researcher at the University of Florida, said the purple shapes resembled mitochondria but did not appear in that form in real cells. Robert Hirst at the University of Leicester said the cells did not look like cilia from patients with the disorder. The decisive clue came from Ian Donovan, a doctoral student at UT Southwestern Medical Center, who found a SynthID watermark in the video. That watermarking system was developed by Google DeepMind for generated content, and Google has recently launched its own A.I. detector based on it.

Nikon reviewed the entry again and then withdrew it. The new winner is Nguyen Nam Nhat of Vietnam, whose video shows a roundworm and the single-celled organism Dileptus. Nikon also says it will revisit its rules and judging process for future contests. Xu pushed back, saying the original footage was real and that A.I. was used only in post-processing to color reconstructed grayscale data, not to create the movie, the cilia, or their movement. Supporters argued that microscopy work always involves processing. But for many scientists, the point is simpler than that: if viewers cannot trust that an image is grounded in actual measurements, the picture is already compromised.