Google has announced its plan to leverage machine learning models to automatically estimate the age of users across its platforms, in a significant move to ensure content is age-appropriate and safe, especially for those under 18. This move aims to enhance child protection online as tech companies face mounting pressure from lawmakers to make platforms safer for young users.
Google's AI system will analyze various data points already associated with a user's account, such as browsing history, the categories of videos watched on YouTube, and the age of the account, to determine if a user is under 18. For instance, younger users might search for educational content or cartoons, while older users may engage with more mature topics. The initiative follows similar moves by other tech giants like Meta, which began testing AI systems to identify teens who may have lied about their age.
If Google's model estimates a user is a minor, several protections will be automatically enabled. These include turning on SafeSearch to filter explicit results, restricting sensitive ad categories, and disabling Timeline in Maps. On YouTube, digital wellbeing tools like break reminders will be activated, and content recommendations that may not be appropriate will be limited. Furthermore, access to age-restricted apps on the Google Play Store will be prevented.
In cases where the system incorrectly flags an account as underage, the user can provide age verification in the form of a government-issued ID, a credit card, or a selfie. This multi-option approach has already been used by other companies and in various markets, including Europe. Google emphasizes that its age estimation approach does not involve collecting additional data. The move is part of a broader commitment to comply with child protection regulations like the COPPA in the US and the Digital Services Act in Europe.
The company is also enhancing parental controls through its Family Link app, allowing parents to manage call and message notifications on their child's devices. The initial rollout of the system will begin in the US for a small set of users, with plans for a wider expansion over time.