Google Has New Measures to Combat Explicit Deepfakes

Google has taken significant steps to address the disturbing issue of explicit deepfakes, particularly those involving celebrities and other individuals without their consent. Just a few weeks ago, a search for “deepfake nudes Jennifer Aniston” on Google returned numerous results claiming to have explicit, AI-generated images of the actress. Today, those results have been eliminated.

New Ranking Adjustments

Emma Higham, a Google product manager, announced that the company has implemented new ranking adjustments aimed at reducing the visibility of fake explicit images. These changes have already decreased exposure to such content by over 70% for searches involving specific individuals. Instead of showing problematic results, Google now promotes news articles and nonexplicit content. For example, a search for Aniston’s deepfakes now brings up articles like “How Taylor Swift’s Deepfake AI Porn Represents a Threat” and warnings about “deepfake celebrity-endorsement scams” from sources like the Ohio attorney general.

A Focus on Societal Impact

“With these changes, people can read about the impact deepfakes are having on society, rather than see pages with actual nonconsensual fake images,” Higham stated in a recent blog post. This shift aims to inform the public about the dangers of deepfakes instead of inadvertently spreading harmful content.

Background and Advocacy

These changes follow a WIRED investigation that highlighted Google’s past reluctance to implement several proposed measures to tackle the issue of intimate images spreading online without permission. While Google has made it easier to request the removal of unwanted explicit content, advocates for victims have been pushing for more proactive measures. The challenge for Google has been to strike a balance between regulation and maintaining access to legitimate adult content.

The Rise of AI-Generated Images

The increasing availability of AI image generators, which often come with few usage restrictions, has led to a surge in nonconsensual explicit imagery (NCEI). These tools allow almost anyone to create fake explicit images of others, whether they are classmates or celebrities.

New Measures in Place

Google’s latest measures to combat explicit deepfakes include:

1. Preventing Duplicates: After honoring a takedown request for a sexualized deepfake, Google will now also try to keep duplicates from appearing in search results.
2. Filtering Related Queries: Explicit images will be filtered out from results in queries similar to those mentioned in takedown requests.
3. Demoting Noncompliant Websites: Websites that receive a high volume of successful takedown requests will be demoted in search rankings.

“These efforts are designed to give people added peace of mind, especially if they’re concerned about similar content about them popping up in the future,” Higham wrote.

Ongoing Challenges and Criticisms

While these measures mark a positive step, Google acknowledges they are not perfect. Critics, including former employees and victims’ advocates, believe the company could do more. Although Google warns users in the US searching for explicit images of children that such content is illegal, no similar warnings exist for adult sexual deepfakes. Despite laws against sharing NCEI, Google has confirmed that this policy will not change.

Google’s efforts to crack down on explicit deepfakes reflect a growing awareness and response to the misuse of AI technology. By promoting informative content and demoting harmful websites, Google aims to mitigate the impact of nonconsensual explicit imagery and provide users with a safer online experience. As AI technology continues to evolve, so too must the strategies to protect individuals from its potential harms.