The frightening truth about non-consensual deepfake pornographic images
Google and Bing are facing criticism for displaying non-consensual deepfake pornographic images at the top of some search results.
12punto
A test conducted by NBC News reveals just how vulnerable search engines are when it comes to deepfake pornography.
According to details covered by Chip, search engines Bing, Google, and DuckDuckGo are facing a barrage of criticism for placing non-consensual, AI-generated deepfake pornography at the top of certain search results. These criticisms follow the publication of an NBC News report highlighting concerns regarding the accessibility of AI-driven pornography.
DEEPFAKE HAS GOTTEN OUT OF HAND
Obscene deepfake videos are created using original pornographic material by replacing the faces of the performers (usually actresses) with those of specific real people (such as celebrities). This replacement is performed automatically by artificial intelligence, and it is sufficient for the user creating the fake images to simply feed the AI with photos of the victim.
According to tests conducted by NBC News on a sample of 36 female celebrities using a combination of the performer's name and the word “deepfake,” Bing and Google displayed deepfake videos and images at the top of search results in almost all cases. Google provided 34 such results at the top, while Bing provided 35. NBC News noted that DuckDuckGo also has similar issues, though it did not specify the scale of the problem.
Furthermore, NBC News complained that articles and news reports regarding the issue of non-consensual deepfake pornography were ranked lower than inappropriate content when searching for the term “fake nudes.” However, it is worth noting that NBC News admitted to performing the searches after disabling the safety features on Bing and Google that prevent the display of obscene content to the user.
Despite this minor detail, it can easily be said that the easy accessibility of AI-powered pornography is a significant problem and that anyone searching for such content would likely disable built-in protective features. Non-consensual obscene deepfake content is inherently illegal and therefore falls among the types of content that Microsoft, Google, and others must address.
WHAT CAN BE DONE AGAINST DEEPFAKES?
The easiest way, or perhaps the first temporary step, to implement more advanced technological measures could be to remove websites known for publishing such content from search results or to ban them entirely. As reported by NBC News, since more than half of the top results in their tests consisted of links to a popular deepfake website or its competitors, such a step could significantly reduce the currently harmful nature of search results.
Microsoft was the only company that did not respond to NBC News' request for comment. A spokesperson for DuckDuckGo stated that its main source for web links and image results is Bing. Additionally, DuckDuckGo offers ways to submit privacy-related requests, such as asking for specific content to be removed from search.
Google also offers a similar option for deepfake victims, but NBC News, citing experts and activists, argues that this is insufficient. Furthermore, it is possible that victims who are asked to submit URLs containing sensitive content for review could suffer further harm during this process.