The tech industry’s unofficial motto for two decades was “move fast and break things”. It was a philosophy that broke more than just taxi monopolies^[1] or hotel chains^[2]. It also constructed a digital world filled with risks for our most vulnerable.

In the 2024–25 financial year alone, the Australian Centre to Counter Child Exploitation received nearly 83,000 reports^[3] of online child sexual exploitation material (CSAM), primarily on mainstream platforms – a 41% increase from the year before.

Additionally, links between adolescent usage of social media and a range of harms have been found^[4], such as adverse mental health outcomes, substance abuse and risky sexual behaviours. These findings represent the failure of a digital ecosystem built on profit^[5] rather than protection.

With the federal government’s ban on social media accounts for under-16s taking effect this week, as well as age assurance^[6] for logged-in search engine users on December 27^[7] and adult content on March 9 2026^[8], we have reached a landmark moment – but we must be clear about what this regulation achieves and what it ignores.

The ban may keep some children out (if they don’t circumvent it), but it does nothing to fix the harmful architecture awaiting them upon return. Nor does it take steps to modify the harmful behaviour of some adult users. We need meaningful change toward a digital duty of care^[9], where platforms are legally required to anticipate and mitigate harm.

The need for safety by design

Currently, online safety often relies on a “whack-a-mole” approach: platforms wait for users to report harmful content, then moderators remove it. It is reactive, slow, and often traumatising for the human moderators^[10] involved.

To truly fix this, we need safety by design. This principle demands that safety features be embedded in a platform’s core architecture. It moves beyond simply blocking access, to questioning why the platform allows harmful pathways to exist in the first place.

We are already seeing this when platforms with histories of harm^[11] add new features – such as “trusted connections” on Roblox^[12] that limits in-game connections only to people the child also knows in the real world. This feature should have existed from the start.

At the CSAM Deterrence Centre, led by Jesuit Social Service in partnership with the University of Tasmania, our research^[13] challenges the industry narrative that safety is “too hard” or “too costly” to implement.

In fact, we have found that simple, well-designed interventions can disrupt harmful behaviours without breaking the user experience for everyone else.

Disrupting harm

One of our most significant findings comes from a partnership with one of the world’s largest adult sites, Pornhub. In the first publicly evaluated deterrence intervention, when a user searched for keywords associated with child abuse, they didn’t just hit a blank wall. They triggered a warning message and a chatbot directing the user to therapeutic help.

We observed a decrease in searches for illegal material, but also more than 80% of users who encountered this intervention did not attempt to search for that content on Pornhub again in that session^[14].