I'm convinced we don't have to choose between protecting children and protecting privacy. With careful, privacy-preserving design, such as detection on the device, no data retention and transparent, independently tested systems, we can do both.

Online child sexual abuse has become a pandemic and millions of children are affected every year. It is a public health crisis and the only way we are going to stop it and protect millions more children is if devices block access to child sexual abuse images and videos at source. It is technically possible, and the time has come for the tech companies who make billions from the sale of their devices to be mandated by legislation to do it.

Law enforcement cannot be the only line of defense against child sexual abuse material. Investigation and prosecution remain essential, but they are inherently reactive. If technology can prevent CSAM from being stored, accessed or shared at the device or operating-system level, then prevention must be part of the solution.

The harm may be hidden behind a screen, but it is not distant. It is real harm, inflicted on real children, in real time. Numbers are not just data. They are proof of what happens when prevention is neglected, when systems know harm is happening, and choose not to act. Every day that prevention is delayed, more children are exposed, more abuse is recorded, and more survivors are forced to live knowing that their images still exist online.

Technology has changed the scale of harm, but it has also changed what is possible in prevention. The tools exist to detect and disrupt abuse material, to block livestreamed exploitation, and to stop the money that funds it. What is missing is not technology. It is courage. The standard we walk past is the standard we accept. Child protection is not a niche concern. It is a measure of who we are. Governments, technology companies, communities and every one of us has a role to play.

This report demonstrates the feasibility and urgency of device-side prevention of online child sexual abuse. By making abuse harder to perpetrate and abusive material harder to access, device-side prevention has the potential to fundamentally undermine the business model of online child sexual exploitation.

For decades, it has been far too easy for offenders to sexually abuse and exploit children online. This report presents a solution to change that through on-device child sexual abuse material blocking. By disrupting the ability to create, livestream and view abusive material, this solution has the potential to protect children from sexual exploitation and abuse.

Almost all child sexual abuse content now, including livestreaming, is being created, shared, and viewed on mobile devices, while computers remain a vector. It is a problem that has only grown as technology has become more advanced and available, but now we have an opportunity to leverage those advances for good. I believe widespread deployment of CSAM blocking at device level could be the single most disruptive and effective action we could take to tackle online child sexual abuse.

Apple and Google already run image classifiers at the operating-system level, on-device, today. A phone can recognise material before it displays it, without breaking encryption. What remains is for the companies that control the operating systems to build that recognition to stop the display of abuse material, beneath the apps, where no app can step around it.

I oppose encryption backdoors. This proposal doesn’t suggest it is the path forward. A classifier that runs on device, before an image is encrypted, or after it is decrypted, does not tamper with encryption keys or secure channels. The phone’s own camera software already processes every frame the same way. Stopping the display of the abuse of a child will disincentivise wide-scale production and sharing of this material, lowering the amount of abuse children are experiencing globally.

Protecting children and protecting privacy should not be viewed as competing goals. This is an interesting paper that presents a privacy-conscious approach to blocking harmful content at the device level, while aiming to keep users' data on their own devices and preserve end-to-end encryption. It is encouraging to see the extent to which the proposal engages with privacy, security and accountability considerations. As the paper acknowledges, independent scrutiny, transparency and ongoing oversight are essential for any OS-level blocking mechanism. I believe the next step is establishing robust technical and governance frameworks that provide confidence in what these systems can recognise, what they cannot recognise, and who controls those decisions.