UK Technology Companies and Child Protection Agencies to Test AI's Capability to Generate Abuse Images

Technology companies and child safety agencies will be granted authority to assess whether artificial intelligence tools can produce child abuse images under recently introduced UK laws.

Significant Increase in AI-Generated Harmful Material

The announcement came as revelations from a protection watchdog showing that cases of AI-generated child sexual abuse material have increased dramatically in the past year, rising from 199 in 2024 to 426 in 2025.

New Legal Framework

Under the amendments, the authorities will allow designated AI companies and child safety organizations to inspect AI systems – the underlying technology for conversational AI and visual AI tools – and ensure they have adequate protective measures to stop them from creating depictions of child sexual abuse.

"Fundamentally about stopping exploitation before it occurs," declared the minister for AI and online safety, noting: "Experts, under rigorous protocols, can now identify the danger in AI systems early."

Tackling Legal Obstacles

The amendments have been introduced because it is against the law to produce and own CSAM, meaning that AI developers and others cannot generate such images as part of a testing regime. Until now, officials had to delay action until AI-generated CSAM was uploaded online before addressing it.

This legislation is designed to preventing that problem by enabling to stop the production of those images at their origin.

Legal Framework

The changes are being added by the authorities as revisions to the criminal justice legislation, which is also implementing a ban on possessing, creating or distributing AI systems designed to create child sexual abuse material.

Practical Impact

This recently, the minister visited the London headquarters of a children's helpline and heard a mock-up conversation to advisors involving a report of AI-based exploitation. The interaction depicted a adolescent requesting help after being blackmailed using a sexualised AI-generated image of himself, created using AI.

"When I learn about children facing blackmail online, it is a source of intense frustration in me and justified anger amongst families," he said.

Alarming Statistics

A prominent online safety foundation reported that cases of AI-generated exploitation content – such as webpages that may include multiple images – had significantly increased so far this year.

Cases of the most severe content – the gravest form of exploitation – increased from 2,621 visual files to 3,086.

• Girls were overwhelmingly targeted, making up 94% of illegal AI depictions in 2025
• Depictions of newborns to toddlers increased from five in 2024 to 92 in 2025

Sector Response

The law change could "constitute a vital step to ensure AI tools are safe before they are launched," stated the chief executive of the internet monitoring organization.

"AI tools have made it so survivors can be victimised repeatedly with just a simple actions, giving offenders the ability to create possibly limitless quantities of advanced, photorealistic exploitative content," she added. "Material which additionally exploits survivors' trauma, and makes young people, particularly female children, more vulnerable on and off line."

Counseling Session Information

Childline also released details of counselling sessions where AI has been mentioned. AI-related risks mentioned in the sessions include:

• Employing AI to rate weight, body and looks
• Chatbots dissuading children from consulting trusted adults about harm
• Facing harassment online with AI-generated material
• Online extortion using AI-manipulated pictures

During April and September this year, Childline delivered 367 support interactions where AI, conversational AI and associated terms were discussed, four times as many as in the same period last year.

Half of the references of AI in the 2025 interactions were related to mental health and wellbeing, including utilizing AI assistants for assistance and AI therapy applications.