AI Companions: Community Reflections and Multistakeholder Recommendations

Key Findings

• — Proposes a four-guardrail governance framework for AI companions: preventing emotional substitution (positioning companions as augmentative rather than relational replacements), prohibiting emotional manipulation for engagement/profit, enforcing privacy-by-default for intimate conversational data, and closing accountability gaps across the AI development-deployment lifecycle
• — Survey respondents spanned academia (17.8%), tech industry (16.9%), civil society (14.9%), students (6.9%), AI startups (6.9%), and government (4%)
• — Builds on the organization's earlier identification of six categories of concern with AI companions: emotional/psychological impact, human relationships and social skills, privacy and data security, safety and user vulnerability, credibility/trust/transparency, and ethical/business-model conflicts

Methodology Notes

Non-scientific community survey (n=108) combined with a multidisciplinary expert working group process; not a controlled study or systematic review. Self-published by the organization.

Sources