Do AI Abuse Protections Actually Work? Researchers Set Out to Find the Evidence
Generative AI has sharply lowered the cost of producing and spreading image-based sexual abuse material, including “nudification” tools and synthetic sexual images of children. Governments, technology platforms, and app stores have responded with new laws, policies, and enforcement measures. Despite these efforts, Heinz College researchers Shweta Singh and Michael D. Smith identify a critical evidence gap: which interventions truly reduce harm, whether abuse is displaced elsewhere, and whether protections reach those most at risk—or instead widen existing racial and socioeconomic inequalities.
The UKAI report Young People and AI: The Opportunities and Risks was launched at the UK Parliament on July 15, 2026, by Kanishka Narayan MP—since elevated to the UK’s first cabinet-level AI minister—and Jess Asato MP, alongside Heinz College Visiting Assistant Professor Shweta Singh. The parliamentary launch brought the report’s findings and recommendations directly to policymakers confronting a defining challenge: how to harness AI’s opportunities for young people while protecting them from new—and unevenly distributed—forms of harm.
The report grew out of a UKAI roundtable held at the House of Lords and chaired by Baroness Thangam Debbonaire. Bringing together young people, policymakers, business leaders, academics, technologists, and representatives of nonprofit organizations, the discussion examined how AI is reshaping education, employment, well-being, identity, and online safety.
A central research contribution to the report is jointly authored by Shweta Singh, Visiting Assistant Professor at Carnegie Mellon University’s Heinz College of Information Systems and Public Policy, and Michael D. Smith, Professor at Heinz College. Their proposed research centers on three connected questions:
- Whether legal, platform, and app-store interventions measurably reduce AI-enabled image-based sexual abuse.
- Whether enforcement displaces abuse to substitute tools, channels, platforms, or jurisdictions.
- Whether broadly applied protections narrow—or unintentionally widen—racial and socioeconomic inequalities.
The stakes of this evidence gap are distributional as much as aggregate: an intervention can reduce harm on average while still failing the young people most at risk, because school resources, reporting friction, institutional responses, and access to takedown and support remedies vary sharply across racial and socioeconomic lines.
The wider report underscores this unequal landscape: although approximately 77% of UK teenagers reportedly use AI, children from economically disadvantaged households are about 50% less likely to use it than their more affluent peers.
“Protecting young people in the AI age requires evidence-led regulation and enforcement, not good intentions alone," said Shweta Singh, visiting assistant professor at Carnegie Mellon's Heinz College. "We must identify which interventions reduce harm, whether abuse is displaced to substitute tools or channels, and ensure protections don’t widen racial and socioeconomic inequalities.”
The researchers propose the first causal evaluation of whether legal, platform, and app-store interventions genuinely reduce AI-enabled image-based sexual abuse — or merely displace it — and whether protections work equitably for young people ages 5 to 25, so that mitigation strategies reduce, rather than reproduce, existing disparities.
“In many areas of technology policy, the incentives that maximize firm profits diverge from what best serves society—particularly its most vulnerable members," said Michael D. Smith, professor of information technology and public policy at Heinz College. "When such gaps arise, it is the responsibility of legislatures to realign private incentives with the public good. The research proposed in this volume aims to provide rigorous evidence to support lawmakers as they confront these difficult policy choices.”
To make the research immediately useful to decision-makers, Singh and Smith propose translating the eventual findings into four practical resources:
- Regulator Enforcement Playbook: guidance on interventions that demonstrably reduce harm.
- Platform/App-Store Enforcement Memo: evidence for technology companies and distribution platforms.
- Equity-Centered School Response Playbook: guidance that prioritizes safeguarding and support while avoiding unequal disciplinary outcomes.
- Remedy Access Brief: recommendations to reduce reporting barriers and improve takedown and support success.
###
Summarized from Young People and AI: The Opportunities and Risks, a UKAI report. The report’s research section, “What Works to Reduce AI-Enabled Image-Based Sexual Abuse: Evidence on Displacement and Inequality,” was jointly authored by Singh, S (Carnegie Mellon University) and Smith, MD (Carnegie Mellon University). Copyright 2026. All rights reserved.
About Heinz College of Information Systems and Public Policy
The Heinz College of Information Systems and Public Policy is home to two internationally recognized schools: the School of Information Systems and Management and the School of Public Policy and Management. Heinz College leads at the intersection of people, policy, and technology, with expertise in analytics, artificial intelligence, arts & entertainment, cybersecurity, health care, and public policy. The college offers top-ranked undergraduate, graduate, and executive education certificates in these areas. Our programs are ranked #1 in Information Systems, #1 in Information and Technology Management, #8 in Public Policy Analysis, and #1 in Cybersecurity by U.S. News & World Report. For more information, visit www.heinz.cmu.edu.