IWF data shows AI child abuse imagery surging across Europe

These numbers are a grim reminder that generative models are already transforming abuse at the long tail, not just productivity at the office. The IWF is clear that its 6,310 AI‑generated images in H1 2026 and the 3,443 AI‑generated videos identified in 2025 are operational detections, not full‑web estimates, but the growth rate is striking.([nationaltechnology.co.uk](https://nationaltechnology.co.uk/Internet_Watch_Foundation_Reports_Over_40pc_Rise_In_AI_Generated_Child_Sexual_Abuse_Content.php)) As tools get more powerful and easier to run locally, the cost of producing highly realistic child sexual abuse material collapses towards zero.

For the AGI race, this is an uncomfortable but important datapoint: even pre‑AGI systems at today’s capability levels can already generate harm that is hard to monitor, hard to detect, and devastating in impact. It strengthens the argument for capabilities‑linked regulation that explicitly considers downstream misuse, not just alignment in the lab. It also highlights the tension between Europe’s strong privacy regime and the need for providers to hash, scan and sometimes inspect user content at scale.

Politically, the report is clearly timed to push EU institutions toward agreement on the long‑debated Child Sexual Abuse Regulation. If that deal empowers platforms to proactively scan for AI‑generated abuse, we may see similar mandates applied to other categories of AI content as models get stronger.