Hash-matching remains the foundation of known-content detection, effective at scale and essential. But generative AI has created a different problem. Novel CSAM and image-based abuse, with no prior hash and no existing fingerprint, is now being produced at a volume that reactive detection workflows were not designed to address. Addressing it requires combining hash-based identification of known material with detection capability that can assess content without a prior match, screening at the point of upload before distribution occurs.
If you missed the third and final webinar in our Online Safety Regulatory Readiness Series, you can now watch the on-demand replay. Hosted by Resolver Trust & Safety’s Director of Product, Frances McAuley, the panel brought together Henry Adams, Director of Trust & Safety at Resolver; Hannah Swirsky, Head of Policy and Public Affairs at the Internet Watch Foundation (IWF); Charlotte Aynsley, Head of Online Safety Policy at SWGfL (StopNCII.org); and Brian Levin, Chief Customer Officer at Reality Defender.
Designed for Trust & Safety leaders, detection program leads, and senior practitioners, watch the replay to learn:
- The scale of the unknown content problem. The IWF recorded a 380% rise in actionable AI-generated CSAM reports in 2024, driven by a shift from redistribution of existing material to production of new content. Learn what that means for detection programs built around known fingerprints.
- The three-step detection sequence. Known detection, unknown detection, and proactive pre-distribution intervention are distinct capabilities, not a single continuum. Understand why each requires different infrastructure, mandate, and operating conditions.
- How platforms align mandate, expertise, and access to build unknown CSAM detection capability. Extending detection beyond catalogued content depends on those three conditions. See what a governed operating model looks like in practice and where platforms most commonly fall short.
- The regulatory and legislative direction of travel. Under the UK Online Safety Act and EU Digital Services Act, platforms are already expected to identify emerging risks and demonstrate how detection measures operate. Find out where civil society and government pressure for proactive obligations is heading, and what that signals for program design now.
- Why detection is not the whole answer. SWGfL’s Revenge Porn Helpline recorded nearly 25,000 reports in 2025, yet only 4% of victims go on to report to the police. Hear what a comprehensive response requires beyond image identification, and why fast, tech-first interventions like hash-matching remain critical to reducing harm.
You’ll leave with a clearer understanding of how the detection landscape is changing, what proactive capability requires, and what your platform needs to build before regulatory expectations shift further. Watch the on-demand session above to hear directly from Resolver, the Internet Watch Foundation, SWGfL, and Reality Defender on why synthetic and AI-generated CSAM and NCII requires detection capability that goes beyond what most platforms have built.
 |
Detect what hash-matching can’t.Addressing unknown and synthetic CSAM requires detection capability that can assess content without a prior match, aligned with the mandate, expertise, and access required to operate defensibly. For a deeper look at how Resolver supports platforms building that capability, see our Athena service overview. |
