Analysis of the Proliferation of AI-Generated Child Sexual Abuse Material

Apr 27, 2026, 09:01

Introduction

The Internet Watch Foundation (IWF) has released an annual report detailing a substantial increase in the production and distribution of AI-generated child sexual abuse imagery (CSAM) between 2024 and 2025.

Main Body

Quantitative data provided by the IWF indicates that reports of realistic AI-generated CSAM rose from 193 in 2024 to 491 in 2025, representing an increase exceeding 150 percent. A more pronounced escalation was observed in video content, where the number of instances rose from 13 in 2024 to 3,443 in 2025. This growth is attributed to the emergence of nudifying bots and the utilization of text-to-video and image-to-video technologies. The IWF further notes that such content is appearing on mainstream social media advertisements and AI companion platforms, distributed across both the clear and dark webs. From a technical and forensic perspective, the IWF observes that AI-generated imagery is more frequently classified as Category A (the most severe) than non-AI content, although 47 percent of criminal AI images from the last two years were categorized as Category C. The foundation asserts that these materials often incorporate the physiological characteristics of actual victims, either through direct modification of existing abuse content or via training data, thereby extending the harm to real children. Regarding the regulatory framework, the Online Safety Act, implemented in March of the previous year, mandates that social media entities identify and remove CSAM. However, stakeholders such as Ian Russell have expressed the view that the legislation lacks sufficient ambition to protect minors. While the UK government has proposed allowing designated authorities to scrutinize AI models and intends to criminalize the possession of AI tools and manuals designed for generating CSAM, the IWF maintains that a legal vacuum exists concerning pre-deployment safety testing. Consequently, the foundation advocates for a 'safety by design' mandate for technology developers.

Conclusion

The current situation is characterized by a rapid increase in the volume and sophistication of AI-generated CSAM, prompting the UK government to expand criminal penalties while the IWF continues to advocate for mandatory industry safety standards.

Vocabulary Learning

emergence (n.)

The process of coming into existence or becoming apparent出現

Example:The emergence of nudifying bots has accelerated the creation of illicit content.

forensic (adj.)

Relating to the application of scientific methods to investigate crimes法醫的

Example:Forensic analysis can trace the origins of AI‑generated images.

pre‑deployment (adj.)

Before the deployment or release of a system or product部署前的

Example:Pre‑deployment safety testing is crucial for new AI models.

proliferation (n.)

Rapid increase or spread of something, especially in large numbers擴散

Example:The proliferation of AI‑generated CSAM has alarmed regulators worldwide.

utilization (n.)

The act or process of using something effectively利用

Example:The utilization of text‑to‑video technologies enables realistic imagery.

Sentence Learning

The IWF further notes that such content is appearing on mainstream social media advertisements and AI companion platforms, distributed across both the clear and dark webs.

Reduced Relative Clause: This sentence contains a reduced relative clause 'distributed across both the clear and dark webs' that modifies 'content', omitting the relative pronoun and verb.簡化關係子句: 句中使用了簡化關係子句 'distributed across both the clear and dark webs' 修飾 'content'，省略了關係代詞和動詞。

From a technical and forensic perspective, the IWF observes that AI-generated imagery is more frequently classified as Category A (the most severe) than non-AI content, although 47 percent of criminal AI images from the last two years were categorized as Category C.

Comparative Clause: The clause uses a comparative structure 'more frequently classified as Category A than non-AI content' to compare two classifications, and includes a concessive clause 'although 47 percent...'.比較結構: 句中使用比較結構 'more frequently classified as Category A than non-AI content' 對兩種分類進行比較，並包含讓步子句 'although 47 percent...'。

While the UK government has proposed allowing designated authorities to scrutinize AI models and intends to criminalize the possession of AI tools and manuals designed for generating CSAM, the IWF maintains that a legal vacuum exists concerning pre-deployment safety testing.

Complex Conditional: This sentence features a 'while' clause acting as a complex conditional, juxtaposing the government's proposals with the IWF's assertion.複合條件句: 句中使用 while 子句構成複合條件句，將政府的提議與 IWF 的主張並列。

This growth is attributed to the emergence of nudifying bots and the utilization of text-to-video and image-to-video technologies.

Nominalization: The nouns 'emergence' and 'utilization' are nominalizations of verbs, turning actions into abstract concepts.名詞化: 句中的 'emergence' 和 'utilization' 為動詞的名詞化，將動作轉為抽象概念。

Quantitative data provided by the IWF indicates that reports of realistic AI-generated CSAM rose from 193 in 2024 to 491 in 2025, representing an increase exceeding 150 percent.

Participial Phrase: The participial phrase 'representing an increase exceeding 150 percent' functions as an adjective modifying 'rose', providing additional information.分詞短語: 句中的分詞短語 'representing an increase exceeding 150 percent' 作為形容詞修飾 'rose'，提供額外資訊。