What happened
BBC interviewed data‑protection advisers, university professors and computer‑science lecturers who cautioned parents against uploading children’s photographs to AI services. They explained that many platforms can retain images, use them to train models, and expose metadata that reveals personal details. While some services offer settings to opt‑out of training, visibility and control vary. The discussion also referenced a recent Internet Watch Foundation report showing a rise in AI‑generated child sexual abuse imagery in early 2026.
The BBC article quotes George Martin, a security and data‑protection adviser at South West Grid for Learning, who stresses that parents should understand what happens to images after upload. He notes that most AI platforms default to allowing training on uploaded content unless users change settings, and that visibility into these settings varies across services.
University of Bath behavioural‑science professor David Ellis and Cardiff School of Technologies lecturer Dr Simon Thorne echo the concerns, warning that uploaded identifiable images can be stored, leaked, and later used to train future AI models. Thorne advises a general rule of never uploading personally identifiable photos to cloud‑based AI services.
The piece also references a report by the Internet Watch Foundation showing a surge in AI‑generated child sexual abuse images in the first half of 2026, compared with the entire previous year. Martin links this trend to the ease of uploading children’s photos to AI tools, noting the difficulty of removing such data once it is online.
Why it matters
The advice highlights a growing privacy and safety challenge as AI image tools become mainstream. If children’s photos are inadvertently used to train models, they can be repurposed for malicious ends, including synthetic child sexual abuse material. Metadata embedded in school photographs can also reveal location and school information, increasing the risk of targeted exploitation. The issue underscores the need for clearer platform policies, stronger user controls, and public awareness about data permanence in AI systems.
The potential for children’s images to become part of training datasets raises serious ethical and legal questions about consent, privacy, and the creation of harmful synthetic content. This risk is amplified by the fact that metadata embedded in photos can disclose sensitive information such as school names and locations.
The rise in AI‑generated child sexual abuse material reported by the Internet Watch Foundation demonstrates a tangible harm that can stem from lax data‑handling practices. This underscores the urgency for both platform operators and policymakers to implement robust safeguards.
For educators and parents, the guidance highlights a trade‑off between the convenience of AI‑powered tools and the long‑term security of minors’ digital footprints, prompting a reevaluation of how such technologies are adopted in schools and households.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Why can ethical evaluation not be reduced to one model score?
What to watch next
Future regulatory proposals on AI data protection, platform‑level opt‑out mechanisms, and industry standards for handling minors’ images will be closely monitored. Parents and schools may also adopt local, offline AI tools to avoid cloud‑based data collection.
Legislative developments in the UK and EU concerning AI data protection for minors, including potential amendments to the UK’s Online Safety Bill.
Implementation of clearer opt‑out settings by major AI image‑generation services, and whether these settings become default or require user action.
Adoption of local, offline AI solutions by schools and families as a way to mitigate cloud‑based data collection risks.