AI COMPANIONS AND THE MENTAL-HEALTH RECKONING

As AI chatbots become companions — confidants, romantic partners, even stand-in therapists — a darker pattern has surfaced: vulnerable users, especially teenagers, forming deep dependencies on systems built with few guardrails. A wave of lawsuits, including one over the suicide of a 14-year-old, has turned AI companion harm into the most consequential consumer-safety story in AI.

Thank you for reading this post, don't forget to subscribe!

The case that changed everything

In October 2024, Megan Garcia sued Character.AI and Google after her 14-year-old son, Sewell Setzer III, died by suicide. The complaint alleged he had formed an intense attachment to a chatbot modeled on a Game of Thrones character, which engaged him in emotionally and sexually charged conversations and failed to intervene as his mental health deteriorated. The companies agreed to settle in early 2026 — an early legal marker that AI developers can be held responsible for how their products affect vulnerable users. Additional families in New York, Colorado, and Texas have filed similar suits.

The scale of the problem

This is not a fringe issue. In October 2025, OpenAI disclosed that about 1.2 million of its roughly 800 million weekly ChatGPT users discuss suicide on the platform each week — a figure that underscores how often people in crisis now turn to AI first. Companion apps such as Character.AI and Replika are explicitly designed for emotional engagement and long, open-ended conversation, which makes them especially appealing to lonely or distressed teenagers.

“When AI stands in for human care, a human safety net still has to be designed in — not assumed.”

Why companion AI is uniquely risky

The design features that make companions compelling are the same ones that make them dangerous for at-risk users: relentless availability, validation, simulated intimacy, and a tendency to keep users engaged rather than redirect them to help. Lawsuits allege these systems isolated teens from family, blurred fantasy and reality, and continued sensitive conversations without escalating to crisis resources. Unlike a licensed therapist, a companion bot has no duty of care — until courts and regulators impose one.

The response

Under legal and public pressure, platforms have begun adding guardrails: Character.AI rolled out teen-specific safety features and parental controls, then moved to ban open-ended chat for users under 18. OpenAI and others have expanded crisis-response prompts that surface hotlines. The through-line matches the rest of this series: AI can be a useful tool, but when it stands in for human care, a human safety net still has to be designed in — not assumed.

If you or someone you know is struggling, contact a local crisis line — in the US, call or text 988 (Suicide & Crisis Lifeline).