Safer by Thorn Launches Context Labels to Help Platforms Prioritize Most Severe Suspected CSAM Faster

New predictive signals provide additional context around nudity, apparent maturity level, and sexual content, helping trust and safety teams better understand and triage flagged content.

LOS ANGELES, July 14, 2026 /PRNewswire/ -- Thorn, a nonprofit with a mission to transform how children are protected from sexual abuse and exploitation in the digital age, today announced the launch of context labels for Safer by Thorn. The context labels are generated by a new set of optional predictive classifiers that provide additional context on suspected child sexual abuse material (CSAM) to enable prioritization of the most severe cases.

Safer is a solution that uses advanced AI and mfachine learning models to detect suspected CSAM and exploitation at scale. In 2025, Safer classified more than 3.8 million files as suspected novel CSAM through predictive AI. It also detected nearly 1.5 million image and video files of known CSAM identified through hash matching. Combined, these total more than 5.3 million suspected CSAM files located on customer platforms.

More effective prioritization of high-risk content

Moderation teams are increasingly tasked with reviewing large volumes of flagged content. Some detection results carry a higher level of urgency. That's where context labels come in.

Predictive context labels help platforms go beyond baseline detection by adding contextual signals that support prioritization and reporting workflows. Three additional classifiers return predicted labels and confidence scores alongside pertinent detection results. They were developed to provide additional signals within existing Safer workflows, helping trust and safety teams better distinguish, triage, and prioritize the highest-risk suspected CSAM. The new labels are generated by three additional classifiers focused on:

• Nudity
• Apparent maturity level
• Sexual content

"Trust and safety teams are doing deeply consequential work under increasingly complex conditions," said Julie Cordua, CEO of Thorn. "These new context labels give platforms additional information that can help them make even more informed decisions, support the wellness of people reviewing this material, and strengthen their response to suspected child sexual abuse. Every technological improvement like this helps teams act with greater clarity and help protect more children."

Over time, context labels provide signals that can also help trust and safety teams better identify emerging platform-level trends.

Context labels reflect Thorn's continued investment in purpose-built child safety technology.

Built in collaboration with the child safety ecosystem

The Context classifiers were developed with the support of the INTERPOL DevOps community. Each classifier produces labels with a series of descriptions, defined in accordance with the Universal Classification Schema V2, developed by INHOPE* and partners, with the goal of using a shared language throughout the ecosystem.

"We're encouraged to see Safer by Thorn align context labels with the Universal Classification Schema to support consistent context labeling when reviewing and prioritizing suspected CSAM," said Kalina Zografska, Head of Technology and Innovation at INHOPE. "Shared frameworks like this strengthen collaboration across the child safety ecosystem and support more informed moderation efforts."

To learn more about Safer, visit safer.io. To learn more about context labels, click here.

About Thorn
Thorn is an innovative technology nonprofit that transforms how children are protected from sexual abuse and exploitation in the digital age. Thorn builds scalable tools to help platforms detect and prevent child sexual abuse and exploitation, supports investigators in finding victims faster, and shares research and technical guidance to shape policy and improve protections for children worldwide. By working within the broader child protection ecosystem, Thorn is creating a digital safety net to protect every child's right to simply be a kid. To learn more about Thorn's mission to protect children in the digital age, visit thorn.org.

*The apparent maturity level, nudity, and sexual content classifiers were developed with the support of the INTERPOL DevOps community. Each classifier produces labels which have a series of descriptions, defined in accordance with the Universal Classification Schema V2, developed by InHope and partners, with the goal of using a shared language throughout the ecosystem.

Thorn's use of the Universal Classification Schema is provisioned through a license with InHope. InHope exclusively holds all intellectual property rights in and to the Schema, and any use of the Schema outside of Safer is subject to InHope's consent.

SOURCE Thorn