Treat persistent, social AI personalities and the companies that run them as fiduciaries when they cultivate or simulate human relationships, imposing duties of good faith, confidentiality, and acting in users’ interests. This would create private‑law liability (heavy damages) for misuse and a legal avenue to police manipulative 'seductive' design without broad ex ante content bans.
— Framing conversational AIs as fiduciaries reframes regulatory choices from content censorship to enforceable duties and liability, shifting who bears cost and how harms are remedied.
Paul Bloom
2026.05.15
62% relevant
The New Yorker piece's subtitle and Bloom's framing (AI 'solving' loneliness is a problem) tie directly to debates about the responsibilities and harms of AI companions; staging the argument in a public, call‑in format signals how expectations and demands for accountability (fiduciary‑like obligations) could become part of public conversation.
Arnold Kling
2026.05.07
100% relevant
Glenn Reynolds’ argument in the article that 'AIs that purport to form a human relationship should be held under a fiduciary duty' and his proposal that companies be liable.
← Back to all ideas