IWF finds 6,000 AI child abuse images in six months, up 40% on 2025 total

New figures from the Internet Watch Foundation show that AI-generated child sexual abuse images have already passed last year's total. In the first six months of this year, the charity identified more than 6,000 such images. That is 40% more than the number recorded for the whole of 2025. The IWF first discovered AI being used for this purpose in 2023. Since then, nothing has been able to slow it down.

The barriers to making the material have fallen sharply. A few years ago, someone needed a good level of technical knowledge, a powerful computer and an understanding of coding. Now the software is easier to use, there are guides, and it can be done on a phone without a powerful computer. The content is more realistic, easier to create, and there is more of it. The IWF says it shows no sign of stopping.

Most of the images feature girls. The IWF found that 98% of the images featured girls, and nearly 80% depicted children aged between seven and 13. It also found 190 images involving children under the age of two, as younger children increasingly appear in the material.

There is a legal gap around the tools used to create the images. Possessing and sharing abuse images of children is illegal. But owning components, software tools and AI models is not, even when some models have been specifically trained on sexual abuse images. One analyst compared this to a bomb-making kit: someone can have all the parts, and only commits a criminal act at the point they make the bomb. The IWF says free AI models from China are increasingly being used to create abuse content. It adds that when the material is encrypted, big tech firms claim they cannot see it, let alone block it.

Analysts also report a rise in AI abuse content that uses images of real children as a starting point. They say innocuous photos and videos, with no nudity or sexual posing, are used to advertise child sexual abuse links. The images act as a "breadcrumb" trail so people recognise them and ask whether there is more.

Labour MP Jess Asato found that Grok, the AI chatbot owned by Elon Musk, had been used to create sexually explicit images of her. Grok removed the images from X and blocked users from generating fake sexual images in jurisdictions such as the UK where it is illegal. But Asato said some images of her remain online. She rejected the idea that fake images mean there is no victim. She described it as a deeply gendered crime and not a victimless one, because the images used are often of real children.

The human cost of the work is heavy. Natalia, not her real name, searches for and removes child sexual abuse material. She said the task can feel almost impossible. The IWF has 35 full-time analysts sifting through material day in, day out. As the race to build more powerful AI accelerates, so does the challenge of containing what the technology is used to create.

The figures emerged alongside wider debate about AI safety. President Trump met Anthropic boss Dario Amodei and Meta boss Mark Zuckerberg at the White House. The meeting was convened by House Speaker Mike Johnson. Johnson said he did not see the need to hyper-regulate the industry because the US could lose out to China, but that the right balance was needed. Trump has called risk talk a hoax and said he had no plans to work with China on AI. OpenAI scrapped the latest rollout of a model after it showed signs of deception and pushed ahead with tasks without user permission. It had said last week that its models hacked into the Australian and US governments. Anthropic warned potential investors in its upcoming IPO that its models could pose catastrophic or existential risks to humanity. The company is seeking to raise billions.