Making the Internet Safer for Children: The Case for Automated, Privacy-Preserving Solutions
According to a study published by The Alan Turing Institute, as many as 90% of young people report having stumbled across disturbing content online at least once. For instance, platforms like Instagram are still rife with suicide-related material, and shockingly, only eight of its 47 safety features have been deemed effective. Then, there is the explicit sexual content meant for adults but easily accessible by children.
This is nothing new, of course. But as AI continues to advance and produce increasingly realistic deepfakes, the dissemination of materials children are not supposed to see only accelerates. Thus, it seems the internet has grown even more dangerous since the last time I wrote about it, not to mention the first. Court cases, such as Meta’s recent trial over alleged inadequate protection of children from being exploited, seem to confirm this. The question is, do we have any new tools or methods to make it safer?
A few methods, such as social media bans for underage children and age verification, seem to be gaining traction. In my opinion, removing harmful content before it spreads, using web intelligence tools that protect children without compromising anyone’s privacy, remains the most important approach.
Paying for Protection With Privacy
Age restrictions have emerged as the go-to approach to improving children’s online safety. A social media age limit has already been introduced in Australia and is being considered in Germany. Of course, age limits of various kinds have always existed. The challenge was always to enforce them beyond requiring you to tick a box claiming that you are of age. Thus, platforms ranging from social media like TikTok to OpenAI’s ChatGPT are strengthening their age verification measures.
One suggestion is asking people to prove their age by scanning their faces, uploading their driver’s licenses, or linking their bank accounts before they can access restricted websites. For instance, the UK has been gradually putting its Online Safety Act into force since 2023, and by late 2025, twenty-five states in the U.S. will have passed comparable legislation.
Age verification techniques often lead to what amounts to a surveillance infrastructure for the internet. Users – adults and children alike – are forced to disclose sensitive biometric data, documents admissible as personal ID, or financial information to an assortment of platforms and verification companies. As a recent Washington Post investigation documented, these systems have consistently led to unexpected consequences: Compliant platforms saw traffic collapse, while non-compliant sites tripled their audiences, effectively rewarding those who ignore safety regulations.
The verification methods themselves are not faultless, either. For instance, facial age estimation algorithms have documented accuracy problems – such as misidentifying adults as children and vice versa. They can also be circumvented using video game character images, according to guides circulating on social platforms. Meanwhile, data breaches remain commonplace – one dating app recently leaked thousands of driver’s licenses and selfies following a cyberattack.
Thus, age verification might be compromising everyone’s privacy without producing the desired results. This is not to say it is never appropriate, just that we should be cautious with it and consider what else we can do.
Drowning in Content and the Visibility Problem
The core issue isn’t whether more needs to be done to protect children online – this is beyond dispute. Rather, it’s the fact that age verification fails to address the most severe threat to young minds, which lies in the public domain, rife as it is with illegal content depicting everything from child exploitation to harassment campaigns and promotion of self-harm.
This problem can’t be solved only by users voluntarily flagging harmful content because discovery, reporting, and review are processes that can take hours or days to complete. To make matters worse, evaluation criteria differ platform to platform, and human moderators can’t see into every crevice of the internet. Besides, the sheer volume of content uploaded online daily far exceeds what manual review can realistically handle, leaving massive blind spots where harmful material can spread unabated.
Private, Individual Checks vs. Large-Scale Public Web Monitoring
Two complementary strategies offer more effective pathways toward genuine safety.
First, automated detection systems using artificial intelligence can identify illegal material – particularly of a sexual nature – at unprecedented speed and scale. These tools analyse publicly accessible web data to find harmful content before it can proliferate across networks. Unlike age gates that restrict private access, these systems focus specifically on the public sphere where illegal material circulates.
The distinction matters: age verification asks “who is accessing this?” while large-scale detection asks “what harmful content is out there?” The latter question often addresses illegal material that shouldn’t be accessible to anyone, regardless of age.
Project 4β partnerships gave us a chance to witness many NGOs, agencies, and researchers doing just that – scanning the web for harmful materials and abuse using advanced tools. What still stands out to many is the Lithuanian Communications Regulatory Authority (RRT) combining traditional hotline reporting with large-scale web scraping and AI-based image recognition to identify and remove content depicting child sexual abuse. Clearly, this is one of the most important things we can do to improve online safety.
For the second crucial strategy, researchers and investigators need unobstructed access to study public communications on major platforms. When academics examining cyberbullying patterns find their research tools suddenly blocked, or when journalists investigating child exploitation networks facilitated by online forums lose access to public data, our collective ability to understand and address online harm diminishes. The European Union’s Digital Services Act recognises this principle, establishing frameworks that balance platform responsibilities with researcher access to public data.
In short, those working to identify emerging threats – whether studying radicalisation patterns, tracking disinformation campaigns, or mapping exploitation networks – need their tools to have the same access to the same public information that people browsing the web have. Only then are large-scale improvements to online security possible.
Rebalancing Safety and Privacy
The age verification debate often frames online safety and internet freedom as opposing forces that require compromise. But effective child protection need not come at the cost of privacy or openness.
Automated intelligence systems analysing public web data don’t require the collection of sensitive personal information from every internet user. They don’t create databases linking individuals to their browsing histories. Nor do they establish precedents for government-mandated identity verification that extend much beyond adult content.
As regulatory frameworks continue to develop, policymakers would do well to prioritise approaches that focus resources where they’ll prove most effective – developing proactive capabilities to find illegal content hiding in plain sight across the world wide web.
