Generative AI in child sexual exploitation and abuse: views from UK law enforcement

Research published in AI & Society reports that the Internet Watch Foundation documented a 380 percent increase in reports containing AI-generated child sexual abuse material, with UK law enforcement interviews highlighting significant inadequacies and gaps in current safeguards associated with multi-turn AI tools, including interaction-centered misuse patterns and a testing gap between existing AI safety evaluation methods and the real-world practices of offenders. The research notes that the recent UK Crime and Policing Act 2026 expands offenses to address child sexual abuse image generators, AI-generated CSAM, and online activity intended to facilitate abuse, generating a corresponding need for lawful, carefully governed means by which developers, researchers, regulators, and child-safety organizations can test whether systems facilitate such harms. The study emphasizes that child-sexual-exploitation-focused AI safety cannot rely on generic red-teaming alone and requires practitioner-informed, legally authorized, and tightly controlled evaluation protocols encompassing multi-turn grooming scenarios, vulnerability-seeking interactions, and prompt-chaining across systems.

Multi-turn conversational exploitation—progressive trust-building, testing boundaries, eliciting compliance—is precisely the behavioral signature that pattern-level detection is built to recognize. Backed by Startmate and operating as a world-leading AI online-safety platform, Guardii applies real-time analysis to the structure and progression of direct-message exchanges, identifying grooming escalation across multiple interactions, coercive pivots from rapport to demand, and the sustained targeting that the research identifies as inadequately addressed by static content filters or single-prompt evaluation. The anti-grooming detection module surfaces children at risk before exploitation is complete, providing parents, schools, and law enforcement with actionable intelligence grounded in the conversational dynamics that offenders rely on, closing the operational testing gap UK law enforcement practitioners highlighted.