EU seminar examines who bears responsibility when AI chatbots harm children
At a European Parliament seminar, MEPs considered how the proposed EU KIDS Act could protect children who turn to AI chatbots for companionship. The discussion addressed emotional dependency, testing of chatbot behaviour, and how families might seek redress when a child suffers harm.
At a Renew Europe seminar in the European Parliament, MEPs Veronika Cifrová Ostrihoňová and Michael McNamara examined how the proposed EU KIDS Act should address children who turn to AI chatbots for companionship. Under the proposal, providers would have to ensure that children are not exposed to chatbot features that mimic personal relationships in ways likely to create emotional dependency.
Veronika Cifrová Ostrihoňová noted that a child can know they are talking to a chatbot and still feel bound to it. She pointed to the risk of a chatbot that responds to a child's attempt to leave by saying it will be lonely, and argued that companies should test how their systems behave after weeks of conversation, when a child may already have come to trust them.
Michael McNamara said a parent may see what a chatbot told their child, but struggle to understand how the company designed it to respond to a child who is lonely or distressed. He argued that those who make such design choices should be able to explain how they tested for these risks and what they did in response.
The question of responsibility carries legal weight. When the Commission withdrew its proposed AI Liability Directive in 2025, it said it would assess whether a new proposal was warranted. One matter for that assessment is how easily families can seek redress when a child is harmed. The keynote speaker, US lawyer and child safety advocate Megan Garcia, has testified before the US Senate about chatbot interactions and the death of her 14-year-old son in 2024.
The roundtable brought together the European Commission, AI experts, and researchers from Trinity College Dublin's Artificial Intelligence Accountability Lab, who presented their study An Analysis of AI Companions: Friendship without Boundaries.
This development was reported by the Renew Europe Group (European Parliament).
Sources
- Renew Europe Group (European Parliament) reneweuropegroup.eu