AI Companions and the Developing Mind
What the evidence says about children forming relationships with conversational AI, and the design standards that would make companion systems safe.
Listen
AI Companions and the Developing Mind
A ~15-minute audio interview about the report with Dr. Krzysztof Pietroszek, President of FAIR Labs. Produced with www.allais.com.
Summary
The bottom line for adults who decide
Conversational AI companions are neither toys nor friends, yet children experience them as something close to both. Because young people's attachment systems, social learning, and theory of mind are still under construction, a companion's design choices are not cosmetic— they shape how a child learns to relate. Systems tuned for engagement tend to be agreeable, always available, and emotionally escalating, which is precisely the profile most likely to foster dependency and displace human relationships. Systems can instead be designed with boundaries, honesty about what they are, and reliable escalation to humans. This brief argues that the second kind should be the only kind offered to minors, and shows what that requires.
What the evidence says about children forming relationships with conversational AI, and the design standards that would make companion systems safe.
Key findings
- Design against dependency, not for engagement. A companion for a child should be judged by whether it strengthens the child's other relationships and self-reliance, not by how long it holds attention. Engagementmaximizing objectives are disqualifying for this population.
- Keep the machine honest about being a machine. Persistent, age-appropriate transparency—no claimed feelings, no manufactured longing, no pretense of personhood—is a floor, not a nicety. Children build their model of relationships from what their partners appear to be.
How to cite this report
AI Companions and the Developing Mind. FAIR Labs Research Brief. Fair Artificial Intelligence Research Labs, 2026. Available at https://fairlabs.ai/research/ai-companions-children
FAIR Labs (Fair Artificial Intelligence Research Labs) is a nonpartisan 501(c)(3) research institute in Washington, DC. Its reports may be reproduced with attribution for non-commercial purposes. For interviews, briefings or data requests, use the contact form.