The 40% understates the pace

Read the headline number carefully. 6,310 images in six months versus 4,512 in twelve means the monthly rate more than doubled year over year. The IWF describes the increase as 40% over the 2025 full-year total, which is accurate and also the gentlest possible framing. The pace is the story: analysts reached the previous year's whole count before June was over.

There is a second shift worth the same scrutiny. Category C, the least severe classification, grew from 62% of the AI total in 2025 to 88% this year. That looks like the material got less severe, but the IWF's own data does not resolve whether the mix of material changed, whether grading criteria held steady, or whether detection improved unevenly across categories. When a share moves that far, the first question is whether the ruler moved too.

The second count is the uglier one

The 6,310 figure counts generated imagery. The Report Remove number counts real children: 420 of them reported suspected AI-manipulated explicit images of themselves in the first half of the year, already past the 397 of all of 2025. Photorealistic generation means victims cannot tell whether anyone believes the images are real. The IWF hashes confirmed illegal imagery so tech companies can block it, but hashing only covers what has already been identified — it is a rear-view mirror.

And the format everyone is watching worst is the one counted least. The new figures cover still images. Separately, the IWF's 2025 tally put AI abuse videos at 3,443, up from 13 the year before. The steepest growth curve in the whole dataset has the thinnest monitoring.

The government says the law is enough

The IWF wants the opposite. Its head of policy backed parliamentarians calling for binding AI legislation that would compel companies to build models that cannot produce this material — safety by design, not takedowns after the fact. The charity has made the same case to the EU.

The UK government's position, as relayed by spokespeople: existing law already covers the abuse imagery, adapting a model to produce it carries up to five years, and ministers are "going further" by banning AI models used to create the material. AI minister Kanishka Narayan says "nothing is off the table" on fresh regulation, but any proposal faces his test: "Will it make the British people safer?" No new AI-focused bill is imminent.

The standoff is familiar: a charity holding a doubling curve, a government holding a test the curve has not been shown to fail. Meanwhile the curve doubles again.

Sources

  1. [1] Tech News Vision, “Internet Watch Foundation reports over 40% rise in AI-generated child sexual abuse content” (Oct 5, 2026)Read source
  2. [2] metirAI Blog, “IWF: AI-Generated Child Abuse Imagery Hits Record in 2026” (Oct 5, 2026)Read source
  3. [3] AI Understanding, “IWF reports AI-generated child abuse images exceed 2025 total in 2026”Read source
  4. [4] Brit Brief, “AI child abuse images surge 40% in first half of 2026, IWF finds”Read source
  5. [5] ResultSense, “AI child abuse images already outstrip 2025, IWF says”Read source
  6. [6] Tech Times, “AI-Generated Child Abuse Images Are Rising: Here's What Parents Need to Know” (Oct 5, 2026)Read source