Participants discussed the harms, risks, and opportunities throughout the meeting, and their key points are summarised here.
Risks and harms
Given that AI is evolving quickly and increasingly ubiquitous, the risk of AI harms is almost inescapable.
Children and young people are already experiencing multiple harms because of AI, for example: deep fakes are being used to generate sexual abuse material; the boom in agentic AI increases the risks of agents distributing and socialising child sexual abuse material; children are engaging with AI chatbots for mental health support, including in ways that may expose them to unsafe content or harmful guidance; and AI generated images are being used to inflame violence in conflict or post-conflict communities.
AI impacts multiple dimensions of young people’s lives, such as learning, mental health and wellbeing, cognitive development, bodily autonomy, critical faculties, activity displacement, emotional dependency, and data privacy. Worrying signs show a 200% increase in Childline counselling sessions relating to AI. Young women fear ‘nudifying’ tools, deepfakes, and for their own personal safety; some are choosing not to participate in the public realm of politics. Many university students consider their chosen degrees to be worthless in the face of a challenging job market further transformed by AI. Furthermore, harms are not always immediately evident, and broader societal harms can emerge over time.
AI has the potential to increase access to tools and lead to greater empowerment among children and young people; however, AI tools can make inequality worse. This is especially the case when deployers and users do not understand how AI models are trained, the inherent bias of data, and the sources and use of personal data. It is extremely challenging for children, young people, and parents to educate themselves on AI. AI literacy is critically important, yet there is a huge inequality gap between those who have access to understanding and using AI meaningfully and those who do not. Children and young people in lower socio-economic groups who do not have access to education are likely to be at greater risk of harm.
Young people also expressed that while workplaces and employers are using AI in methods to hire people, and in graduate roles, the education system is not quick enough to respond to equip students for their future in a world of AI. The education system at all levels must become fit-for-purpose to respond to AI in a changing world, including improving literacy and accessibility.
Opportunities
Young people expressed disillusionment and distrust in the systems that should protect them from the harms of AI, while tech companies expressed greater positivity around the potential of AI to improve lives and economies. More optimism emerged when campaigners and technologists talked about using AI to ensure that AI is safe.
Participants discussed how AI can be used as a tool to combat harms that children experience online, and how it can be used to disrupt the commercialisation of harms. Safety interventions can sit in many places, such as guardrails in AI, safety at the infrastructure level, in search functions, on platforms and on devices; all offer different opportunities and challenges. Safety tech companies have built AI tools that give children safer online experiences, but they are fighting against the economic model of online platforms financed by advertising.
Contentious issues of privacy, inclusion, and freedom of speech can challenge the overall direction and moral decisions that we need to take as a matter of urgency to protect children from AI-generated harm.
Suggestions for action included:
- Ensure a pro-social element in the principles of AI design.
- Use AI classifiers more widely to detect harms and block the ability to view harmful material.
- Do not just focus on technological solutions, but work to overcome the political and economic barriers that exist in keeping children safe online.
- Gather evidence from existing initiatives and programmes about the harms, risks, and opportunities of AI.
- Create and fund a research institute focused on AI and childhood, which includes academia, tech companies and government inputs.
- Conduct a year-long research project in schools based on level of AI use (high, medium, low, none).
- Run a long-term trial of youth use of AI companions.
- Conduct a long-term longitudinal study on AI and the brain.
- Promote digital literacy in schools, communities, and homes.