Google Takes on Deepfakes

On July 31, Google unveiled its latest updates aimed at combating deepfakes. This initiative highlights the tech giants' response to the proliferation of explicit fake content.

What is a deepfake?

A deepfake is a fake image or video generated by artificial intelligence, using advanced image manipulation techniques to create content that appears real. The term "deepfake" is a blend of "deep learning" and "fake." Deepfakes exploit deep learning algorithms to overlay a person's face or voice onto an existing video, creating scenes that may seem authentic but are entirely fabricated.

These manipulation technologies have become extremely sophisticated, sometimes making it difficult to distinguish between real and fake, even for trained observers. Deepfakes can be used for various purposes, ranging from humor and art to misinformation and defamation. Unfortunately, their most concerning use involves creating non-consensual explicit content, often featuring public figures in pornographic scenarios.

The implications of deepfakes are serious and multifaceted. In addition to invading privacy and damaging the reputation of affected individuals, they pose a major risk to personal respect. For example, deepfake videos can be used to spread false information, manipulate public opinion, or even influence elections.

Google's response to threats

In response to these threats, Google, as a major search engine, plays a crucial role in this fight by updating its systems to better manage these contents. Google's ability to identify and filter deepfakes can help reduce their impact and protect potential victims of these digital manipulations.

Facilitated reporting and removal

Google has committed to significantly simplifying the process of reporting and removing deepfakes from its search results, recognizing the urgency of protecting victims of these digital manipulations. The new measures now allow users to more easily and quickly request the removal of fake explicit images.

When a user identifies a deepfake concerning themselves or another person, they can now report this content through a simplified procedure. Google has implemented an intuitive system that guides the user through the necessary steps to make this request. This process includes submitting a removal request that triggers an automated search by Google for similar content.

A major improvement is Google's ability to proactively detect deepfakes associated with a removal request. Once the request is received, Google's systems analyze the reported image or video to identify copies or similar versions on the Internet. This means that even if a deepfake image has been widely shared and reproduced, Google is committed to searching for and removing all similar occurrences it can find in its results.

Moreover, Google has enhanced user protection by making these reporting tools more visible and accessible. Direct links to report inappropriate or harmful content are now more easily accessible in search results and user account settings. This increased visibility aims to encourage users to report deepfakes and other non-consensual explicit content as soon as they become aware of them.

These measures taken by Google represent an important step towards protecting individuals from deepfake-related abuses. By facilitating reporting and automating the detection and removal of content, Google demonstrates its commitment to using its technological capabilities to combat digital threats and promote a safer online environment for everyone.

Content ranking changes

In response to the proliferation of deepfakes, Google has also modified its search algorithm to penalize sites that distribute this type of content. This update aims not only to reduce the visibility of deepfakes in search results but also to promote more relevant and user-friendly content.

One of the main changes concerns the ranking of explicit content in search results. From now on, when users perform searches including people's names and specifically targeting explicit content, Google will prioritize news articles rather than the explicit images or videos themselves. This strategy aims to divert users' attention from harmful content and provide contextual and educational information on the deepfake phenomenon.

Furthermore, if Google detects that a website systematically violates its rules on explicit content, the search engine can take measures to penalize the overall ranking of that site. This approach aims to discourage websites from distributing deepfakes by reducing their overall visibility, without completely removing them from search results. This progressive sanction allows Google to maintain a balance between user protection and freedom of expression on the Internet.

However, this penalization system is not without challenges. A persistent issue is Google's algorithms' ability to accurately distinguish deepfakes from real explicit content. Although significant progress has been made in this area, there are still classification errors that can lead to the removal of legitimate content or the persistence of deepfakes in search results.

The importance of AI watermarks

These updates to Google's algorithm represent a significant advancement in the fight against deepfakes by reducing their visibility and promoting safer and more reliable content. However, combating these digital manipulations requires continuous vigilance and constant technological improvements to ensure a safer online environment for all users. Moreover, Google's initiative underscores the need for more robust tools to combat deepfakes. A promising solution lies in the development of digital watermarks for AI-generated content. These watermarks would allow tracing the origin of images and videos, facilitating their identification and removal in case of misuse. Google's efforts mark an important step, but the fight against deepfakes will require ongoing collaboration and technological innovation to protect individuals from these violations.

By Brice Matter

____________________________

To stay updated on generative AI news, subscribe to the Gennn Newsletter here!