Over the past year, non-consensual deepfake images, particularly of female celebrities, have proliferated online, presenting a challenge for search engines: even if you’re not looking for the material, searching for a particular name can turn up a surprising number of links to fake explicit photos and videos of that person.
Google is trying to address the issue by updating its ranking system, the company announced in a blog post.
Emma Higham, a product manager at Google, wrote in the post that the ranking updates are designed to reduce blatantly fake content across many searches.
If someone uses the term to look for deepfakes without the consent of a specific individual, the ranking system will instead try to serve up “high-quality, non-explicit content,” such as news articles, if available.
The consequences of creating deepfakes without consent
“These changes will allow people to read about the impact deepfakes are having on society, rather than viewing pages full of fake, non-consensual images,” Higham wrote.
Mashable Lightspeed
Higham said the ranking updates have already reduced the number of explicit images that appear in deepfake searches by 70 percent.
Google also aims to downrank explicit deepfake content, but Higham noted that it can be difficult to distinguish between real, consensual content, such as nude scenes featuring actors, and material generated by artificial intelligence without the actors’ consent.
To spot deepfake content, Google will now factor into its rankings whether a site’s pages have been removed from search under its policies, and sites with a high number of removals due to fake explicit images will now be ranked lower in search results.
Additionally, Google is updating the system that processes requests to remove deepfakes from search without consent, a change that should make the request process easier.
Once victims are able to remove their own deepfakes from Google searches, the company’s system will filter all relevant results from similar searches about the victim, aiming to scan for and remove duplicates of that image.
Higham acknowledged there was “more work to do” and said Google would continue to develop “new solutions” to help people who have been victims of deepfakes without their consent.
Google’s announcement comes two months after the White House called on tech companies to stop the spread of explicit deepfake images.
topic
Google Social Good