Pixiv bans all images indistinguishable from photographs, likely to tackle malicious deepfakes 

Pixiv posted an announcement regarding the revision of their Terms of Use and Guidelines on April 10. With the latest revision, the platform has banned the upload of all real-life/photographic imagery and illustrations/images that are equally realistic to real-life photographs. This appears to be an attempt to prevent the upload of deepfake imagery, but has triggered dissatisfaction among some users of the platform.  

Last year, the Japan-based platform for sharing and viewing illustrations, comics and fiction had made several revisions to their Creator Guidelines. These had been largely related to problems arising from the proliferation of AI generated art. Pixiv had introduced rules aimed at controlling “floods” of untagged AI generated art on the platform and preventing losses to their artists caused by imitation/identity theft and unauthorized use of creators’ works for AI training purposes. 

Until recently, Pixiv had a regulation forbidding the upload of extremely realistic/photographic imagery displaying sexual content. But with the latest update, this regulation has been broadened to include non-sexual photographic (or photographic looking) images as well, as the announcement states, “Pixiv prohibits the posting of images that are clearly indistinguishable from photographs, regardless of whether they have a sexual intent or not.” This is likely a means to prevent AI/deepfake images with malicious intent – such as images that violate another’s personality rights or intentionally spread misinformation. 

However, Pixiv’s revision of regulations has sparked dissatisfaction, particularly among artists who create images with a “real-life” level of detail. The repercussions are not mild either, as one user who had uploaded an AI generated parodical portrait of Draco Malfoy looking like a historical samurai (seen above) reported that their account had been suspended because of the image.