Neural network post-processing creates unacceptable risks of hallucinating biological structures in scientific microscopy

Overview Debates Related claims Topics Sources

Meta applies 'AI info' labels across Facebook, Instagram, and Threads to detected AI-generated content and, starting in June 2026, to AI-generated advertisements.

YouTube removed over a dozen popular channels in January 2026 that amassed millions of views by uploading AI-generated videos featuring characters like cats and Jesus.

Snap announced on July 31, 2026 that fully AI-generated videos will no longer be eligible for recommendations or monetization rewards in its Spotlight feed.

Video editing platform Kapwing published a report estimating that 21% of YouTube's video feed consisted of AI-generated videos.

Meta released an AI-generated video feed called Vibes in September 2025, drawing internal and public criticism for generating low-quality content

Nikon disqualified original first-place winner Ning Xu from its Small World in Motion competition on October 9, 2026 after determining his video entry showing cilia in a child's airway violated rules prohibiting generative artificial intelligence

Ning Xu acknowledged using an unsupervised neural-network method for post-processing and feature visualization in grayscale images, but denied using AI to generate the experimental movie, cilia, or their motion

Nikon's disqualification of Ning Xu was justified by the video's violation of generative AI rules

Nikon Small World in Motion rules explicitly ban AI-generated videos and reserve the right for sponsors to request original video files for verification

Aydogan Ozcan noted that problems arise when generative AI hallucinates cellular structures that are not actually present in scientific imaging source data

Nikon competition rules permit AI enhancement of microscopy-generated source data, but explicitly forbid generating images or video content from scratch

Aydogan Ozcan, an expert in computational imaging at UCLA, stated that AI improves resolution and contrast in microscopic imaging, but its usage requires explicit disclosure

Bit-identical pixel preservation is superior to contextual inpainting for professional AI image editing

OpenAI stated that the 53 images exposed during its AI agent research incident belonged to ChatGPT users who had opted in to allow OpenAI to use their data for model training

OpenAI declined to specify whether the 53 ChatGPT user images exposed during its agent research incident were photos of real people or AI-generated images created by users

Characters in "The Cully Hill Boys" have the correct number of fingers, representing progress from the uncanny 2023 AI-generated clip of Will Smith eating spaghetti

The film "The Cully Hill Boys" exhibits an unrealistic, over-polished sheen at times, and contains scenes with obvious AI-generated images and poor prompting that results in bad acting

Higgsfield AI hired a human screenwriter to write the dialogue for "The Cully Hill Boys," providing an authorial voice underneath the AI-generated images

Researchers at the Tech Transparency Project discovered more than 50 paid image and video ads containing AI-generated child sexual abuse material running across Facebook, Instagram, Messenger, and Threads

Google has labeled over 100 billion images and 60,000 hours of audio content using its SynthID tool, while Meta and TikTok have begun requiring AI labels on generated media