Childlight finds AI-linked child abuse reports rose more than 8,400%
Editorial Team
••📖7 min read
Reports linked to AI and child sexual exploitation rose from about 4,700 in 2023 to more than 400,000 in 2025, Childlight finds. The increase exposes growing threats and weaknesses in reporting systems, but does not measure unique victims or confirmed AI-generated images.
AI-linked reports rose from about 4,700 in 2023 to more than 400,000 in 2025.
Reporting changes may explain part of the increase; submissions are not a count of unique victims or confirmed images.
Childlight calls for safer technology design, clearer classifications and stronger investigative capacity.
Reports linking artificial intelligence to child sexual exploitation and abuse rose by more than 8,400% between 2023 and 2025, according to Childlight, the Global Child Safety Institute hosted at the University of Edinburgh. Its 2026 Into the Light Index puts the increase at 8,411%, from about 4,700 reports to more than 400,000.
The index is being presented on November 16 at a ministerial summit in the Philippines involving representatives from more than 100 governments on ending violence against children. It brings a growing threat into view, but the percentage increase cannot establish how much abuse has increased: Childlight cautions in its explanation of the findings that improved detection and reporting may account for some of the change.
Childlight is a research institute launched at Edinburgh in March 2023 to assemble evidence on sexual exploitation and abuse across online and offline settings. Building on its first index in 2024, this edition focuses on East Asia and the Pacific, Latin America and the Caribbean, and North America. It combines surveys, frontline service records, abuse-material reporting and evidence on protection systems. The Human Dignity Foundation supported the index research, according to the report’s acknowledgements.
Childlight’s comparison of AI-linked CyberTipline submissions. These totals do not count unique victims or confirmed AI-generated images; changes in detection and reporting may explain part of the increase. Source: The Guardian.
The FBI warns that ARTEX can accelerate intrusions, with some reaching privileged access in under three hours. Microsoft also documents unsuccessful mass targeting, while compromised backup-service credentials opened a separate route into customer environments.
Editorial Team
The underlying figures come from the National Center for Missing & Exploited Children, or NCMEC, the US organisation whose CyberTipline receives suspected child sexual exploitation reports and makes them available to law enforcement worldwide. The service began in 1998. US companies must report specified suspected exploitation when they become aware of it; their international user bases give the system a global reach.
NCMEC’s 2025 account describes more than 400,000 reports of exploitation with a generative-AI connection. Within those submissions, more than 182,000 concerned offenders possessing, generating or attempting to generate AI child sexual abuse material, or CSAM. The broader figure therefore cannot be read as 400,000 confirmed AI-generated images.
The distinction does not make the threat hypothetical. Since NCMEC began tracking generative AI in 2023, more than 275 victims of AI CSAM have been identified and more than 158,000 images and videos categorised as AI CSAM, the organisation says. Those cumulative figures measure different things from annual report submissions and should not be added together.
AI can also compound existing abuse. NCMEC describes offenders manipulating known abusive imagery to produce new content and classmates or peers using “nudify” applications against children. Fabricated images can become tools for blackmail, with demands for money or further images. An image need not document a new physical assault to cause harm to an identifiable child.
Separate monitoring supports concern about the material’s growing availability. The Internet Watch Foundation, a UK-based hotline that finds, assesses and works to remove online child sexual abuse material globally, said in an October 5 release that it assessed 6,310 photorealistic AI-generated images meeting the legal definition of child sexual abuse between January 1 and June 30, 2026. That was about 40% more than the 4,512 assessed in all of 2025. These are assessed images, a different unit from CyberTipline submissions, and the comparison covers six months against a full year.
More submissions do not always mean more usable evidence
Technology companies determine much of what becomes visible. NCMEC says just over 300 companies submitted CyberTipline reports in 2025, with five providers accounting for more than 75% of submissions. Changes by a few large reporters can therefore materially change the aggregate.
An outlier examined in Childlight’s supplemental analysis illustrates why classification matters. It identifies a separate batch of roughly 1.1 million reports from Amazon AI Services that contained no actionable offender- or victim-related information, warning against interpreting that batch as a surge in generated abuse material.
An August 27 Senate Judiciary Committee update summarising information from NCMEC provides a further qualification: Amazon’s human review determined that 99.6% of its approximately 1.1 million reports were false positives; 4,376 were confirmed CSAM. The update says Amazon retracted false positives and introduced human review of public-web detections before reporting. Amazon AI Services was reporting suspected content it detected; its name does not establish that the material was AI-generated.
Reporting quality is a wider problem. NCMEC says more than 10% of technology-industry reports in 2025 lacked enough information to determine where the offence occurred. Its annual account also records reporting improvements from Amazon AI Services and other providers in early 2026. The relevant test is whether reports enable investigators to identify a victim, offender or location, rather than how many submissions a company makes.
Childlight’s analyst-availability survey covered 11 jurisdictions in East Asia and the Pacific, 24 in Latin America and the Caribbean, and two in North America. The findings represent participating jurisdictions. Source: Childlight — Into the Light Index 2026 supplemental report.
Classifications and investigative capacity both matter
Childlight proposes distinguishing fully synthetic material, altered images of identifiable children, combinations of real and synthetic content, and stylised AI material. These categories can imply different investigative tasks, even when removal remains necessary. A consistent framework would help separate victim-identification opportunities from other forms of harmful content.
The capacity to act is uneven. In research conducted with INTERPOL and other partners, Childlight surveyed national law enforcement and specialist units. Across 37 responding countries and territories in the report’s three regions, 45.9% said they lacked dedicated online child sexual exploitation analysts. Only 35.1% reported digital forensic capability sufficient to support all online investigations of such abuse within 12 months. These are findings from participating jurisdictions, not estimates for every country worldwide.
Routing creates another obstacle. Across 100 jurisdictions examined in the three regions, 43% received CyberTipline referrals directly and 48% through intermediaries. Interviewees in some jurisdictions receiving reports indirectly described limited visibility into how reports were filtered and said few ultimately reached specialist investigators. Those accounts identify a possible break in the referral chain, rather than proving that every indirect route fails.
The broader evidence base is also incomplete. Across 103 countries and areas examined in East Asia and the Pacific, Latin America and the Caribbean, and North America, only 22 had representative survey data covering both offline and technology-facilitated abuse, according to the study’s reported coverage findings. Children with disabilities, Indigenous children, migrants, very young children and those abused within families were among the groups underrepresented.
Childlight is asking governments to treat sexual exploitation and abuse as a preventable public health issue, fund consistent measurement and strengthen survivor support and prevention. Its recommendations also address commercial incentives through regulation, procurement, taxes or levies, alongside safety measures built into technology from the outset. These are policy recommendations, not measures shown by this index to have reduced AI-related abuse.
Removal outcomes offer a more concrete test than submission volume. NCMEC says its 2025 notices had an average takedown time of 2.6 days. But Childlight’s supplemental report lists 147,904 notices that resulted in removal within 12 days and 19,366 that did not. Those figures cover removal notices generally, not AI material alone. NCMEC sends notices to providers, which handle removal.
Even removal is an incomplete outcome measure. Childlight says existing figures do not establish how long content stays offline, and calls for more research into AI-material removal and repeat circulation. Faster first removals therefore need to be assessed alongside whether harmful material reappears.
Governments and platforms now face linked decisions about prevention, report quality and investigative resources. Consistent AI classifications would make the trend more interpretable; evidence on referrals reaching investigators, victims identified and lasting removals would show whether the response is protecting children. The reported percentage increase answers neither question by itself.
Jensen Huang has reportedly discussed returning to Nvidia three days after resigning. Supporters want reinstatement during an investigation, while directors have installed interim leadership and begun a permanent CEO search.
House China committee chairman John Moolenaar demanded answers from Commerce Secretary Howard Lutnick over delayed restrictions on Nvidia customer Megaspeed. The dispute concerns enforcement, overseas access to advanced chips and commercial influence as President Donald Trump prepares to visit China.
Apple reportedly still aims to unveil iPhone-linked glasses at WWDC 2027 and ship before year-end. Siri control of existing apps is the software bet, but delays cloud the schedule and Meta already has an established eyewear lineup.
Treasury Secretary Scott Bessent urged Nvidia to recruit an outside CEO and explain within 10 days whether suspect chip sales inflated earnings. The board is considering internal and external candidates, while export enforcement also faces scrutiny.
Congressional scrutiny and company investigations are testing what Nvidia knew about Megaspeed, a Singapore cloud operator whose Malaysian subsidiary bought nearly $2 billion of its technology. Megaspeed denies wrongdoing; the available reporting has not conclusively traced suspected chip diversion to China.
Commerce Secretary Howard Lutnick is backing narrower restrictions on Chinese access to overseas AI computing, according to a report examining his former firm’s financing relationships. Bitdeer filings establish Cantor’s role and the scale of its client’s planned expansion, but do not establish improper influence on policy.
Microsoft said services were fully restored on Saturday after a global Microsoft 365 outage disrupted email, healthcare communications and public services. It blamed an earlier network capacity upgrade exposed by a later change, with a full technical review still to come.
Cursor says a key exposed by Fireworks led an attacker to unredacted credentials in historical coding-agent records and then its extension cache. Up to 5,000 users potentially received malicious Claude Code builds. Cursor has moved inference off Fireworks; successful endpoint compromises remain uncounted.
Police are investigating a top tax-accountant exam scorer's admission of AI-assisted cheating, while the government reviews national qualification tests held in 2026 and prepares stronger controls.
An early tally puts Japanese listed companies’ first-half net profit growth at 43.2%, led by manufacturers. Filings show AI component demand alongside currency benefits and non-AI earnings gains, while procurement delays and canceled infrastructure reservations expose the costs of expansion.
TSMC plans to back promising chip startups through a new investment arm, alongside separate commitments to OpenAI and Anthropic. The strategy reaches toward future customers and the AI developers shaping chip demand, but the investment rights remain unclear.
Nidus has raised $30 million across pre-seed and seed rounds to automate wire processing in existing U.S. defense factories. Its founders bring manufacturing and government-deployment experience, but customer access and proprietary data still need to translate into measurable production gains.
Nebuly is funding US and European expansion for its AI business-value platform. Adjustable assumptions and private deployment are part of its pitch, but rivals offer similar measurement tools and verified customer returns remain undisclosed.
Han Won-seok says he used a miniature camera, generative AI and a smartphone in tax accountant and CPA exams. His confession puts scrutiny on exam security after proctors reported no suspicious behavior.
Megaspeed says Alice Huang left on August 30, 2025, and has retained outside lawyers for an internal compliance investigation. Singapore’s inquiry began months before that departure; authorities later found insufficient evidence that Huang had committed a crime, while the investigation into the company continues.
HP’s US listing puts a $132,000 starting price on its ZGX Fury shared AI system. Its 748 GB of coherent memory supports a local-computing pitch, but competing GB300 machines, dedicated power requirements and workload-specific operating costs shape the buying decision.
Apple's new U.S. study will examine how up to 200,000 adults understand and use health answers from OpenEvidence, a clinician-focused AI company. Optional personal health data adds context, but competing chatbots already offer data connections, and the study has yet to demonstrate better health decisions.
A $1.5 trillion valuation underpins the estimate of 4,250 Anthropic employee millionaires. The offering remains unpriced, selling restrictions limit liquidity, and the latest filing shows computing costs complicating the company’s rapid growth.
Supermicro says Rubin racks are shipping and Nvidia claims production at scale, while supplier reporting points to larger deliveries in 2027. Memory constraints and differing forecast scopes leave the pace of the transition unresolved.