Pornographic videos: Google Search and Bing overtaken by deepfake

A new independent study reveals that Google and Microsoft search engines are failing to effectively combat porn sites that use deepfake. Videos and photographs doctored by AI are increasing and, with them, the difficulties for victims.

Led by a researcher who wishes to remain anonymous so as not to be targeted, the study shows that the number of pornographic videos faked and posted online after the use of artificial intelligence is increasing sharply from year to year.

The proliferation of tools allowing the creation and distribution of deepfakes seems to pose great difficulties for search engines with a worrying observation: 50 to 80% of users access them with a simple search.

Deepfakes by the hundreds of thousands

The basic principle of deepfake NCEI (synthetic non-consensual exploitative images), namely tampering with a photo or video to give the people in it a different appearance, is as old as the media themselves. But in recent years and the new accessibility of AI, the phenomenon has progressed exponentially, particularly in the field of pornography. A new independent analysis shows that, in the first nine months of the year, 113,000 deepfake videos were uploaded to the sites analyzed.

This represents an increase of 54% compared to the same period in 2022, and the total number of videos produced over the year is expected to exceed the total of the previous two years. Beyond the harm caused to people whose faces are found without their authorization in adult content, it seems that search engines are having great difficulty coping. The research highlights 35 sites that clearly indicate using deepfakes and 300 others that integrate them in one way or another.

Through an analysis carried out by the SimilarWeb service, the search report shows that 50 to 80% of visits are made from a search on search engines such as Chrome or Bing. The latter reacted, specifying that they are actively working to improve the situation, and that it is possible to report unauthorized content using dedicated forms.

Specialists not optimistic about the evolution of the situation

The prospects for the evolution of the situation around the proliferation of deepfakes, especially when they are pornographic or aimed at setting up blackmail, are not very good if we are to believe the specialists. On the one hand, the problem seems global since with the help of a VPN, the anonymous researcher was able to quite easily access content on five continents. On the other hand, Asher Flynn, associate professor at Monash University in Australia, says demand is growing and legislation cannot keep up.

For the moment, complaints and complaints constitute the main levers to have the videos removed, but that is not enough. According to Henry Ajder, an expert in generative AI, you must first make research as difficult as possible. In other words, if we cannot fight against proliferation, we must try to make the content invisible. This involves in particular a downgrading of certain results or by blocking the sites identified. Elsewhere, like on Twitch, it is now grounds for banning.

“ It’s hard to be truly optimistic, given the volume, the scale of these operations, and the need for platforms that historically haven’t taken these issues seriously to suddenly get on board. », indicates Henry Ajder.

Finally, at the end of the chain, we find the victims, on whom the impact of the distribution of fake or stolen content can be enormous. In addition to the obvious attack on people’s privacy, image and even dignity, research reveals a lasting impact and an increase in the phenomenon of cyberharassment. Asher Flynn explains that his work has highlighted a potential impact on the physical and mental health, employment, family and social life of those affected.

Source : WIRED