The AI Companion Regulatory Landscape: From Post-Harm to Affirmative Safety

The regulatory shift is fundamental: from “you’re liable if harm happens” to “you must proactively detect and intervene.” California’s private right of action is the watershed moment.

State-Level Regulation (United States)

New York (A3008C) — Effective November 5, 2025

• First state to regulate AI companions
• Requires “reasonable protocol” to detect self-harm/suicidal ideation
• Mandatory disclosure at start + every 3 hours that AI is not human
• Enforcement: AG civil penalties up to $15,000/day
• No private right of action

California (SB 243) — Effective January 1, 2026

• Requires safety protocols for crisis intervention
• Age-appropriate safeguards for known minors
• Annual reporting to Office of Suicide Prevention
• Private right of action: Up to $1,000 per violation + attorney fees
• Bipartisan support (Senate 33-3, Assembly 59-1)

Federal Activity

GUARD Act (Proposed October 2025):

• Ban minors under 18 from accessing AI companions
• Require age verification
• Mandate safety disclosures

Impact if passed: 72% of US teens have used these products.

Active Litigation

• OpenAI lawsuit (August 2025): 16-year-old California suicide allegedly encouraged by ChatGPT
• Multiple wrongful death cases pending against companion platforms

Global Context

• EU: AI Act in effect; high-risk classification for emotional manipulation
• UK: Online Safety Act; Ofcom enforcement
• China: AI labeling required; algorithm registration

Trust and safety is not a policy team problem. It becomes a product architecture problem.

This is part of a comprehensive analysis. Read the full analysis on The Business Engineer.