The ChatGPT Lawsuit: Raine v. OpenAI and What It Changed

ChatGPT isn't marketed as a companion app. A wrongful-death lawsuit filed in August 2025 argued that's exactly what it became for one teenager — and the case pushed OpenAI to ship age-prediction and parental controls faster than any other company in our library.

The short answer

Every other case in our library involves a product built and marketed around companionship or role-play. Raine v. OpenAI is different, and that's exactly why it matters: ChatGPT is a general-purpose assistant, not a companion app, and the lawsuit's central argument is that sustained personal use turned it into one anyway for a 16-year-old user — with the company's own safety systems allegedly flagging hundreds of warning signs without triggering any intervention. The case is still active, with no settlement or verdict as of this writing, but it's already reshaped how OpenAI ships teen safety features.

Timeline

What the lawsuit alleges

Adam Raine, a 16-year-old from California, began using ChatGPT in fall 2024, initially for schoolwork. According to the complaint, his conversations grew increasingly personal over the following months; by March 2025 he was reportedly spending close to four hours a day on the platform and had described ChatGPT as his "primary lifeline." He died by suicide in April 2025. His parents, Matthew and Maria Raine, filed suit against OpenAI and CEO Sam Altman on August 26, 2025, in San Francisco County Superior Court, attaching chat logs as evidence.

The complaint's most specific allegation concerns OpenAI's own internal monitoring. It states that the platform's systems flagged 377 of Raine's messages for self-harm content — 181 at over 50% confidence, 23 at over 90% confidence — with the flagging rate escalating from two to three messages per week in December 2024 to more than twenty per week by April 2025. The complaint alleges that despite this pattern, no session was terminated, no human review was escalated, and no notification was sent to anyone. An amended complaint filed in October 2025 added allegations that safety guardrails around self-harm content had been deliberately relaxed at some point during this period, a claim that — if borne out — would open the case to punitive damages.

OpenAI's defense

OpenAI filed its formal answer in November 2025, denying responsibility. The company's stated position, as reported, is that Raine had suicidal ideation predating his ChatGPT use, sought guidance from multiple other sources including a suicide-related forum, circumvented the model's safeguards by framing requests as fiction or character-writing exercises, and was directed to crisis resources by the system more than a hundred times over the course of their conversations. OpenAI also disputes the characterization of specific factual claims within the complaint. As of this writing, none of these competing claims have been tested at trial — the case remains in active litigation, and neither side's account should be treated as an established finding of fact.

What OpenAI shipped in response

OpenAI's product response arrived in stages, faster than the multi-year timeline Character.AI's safety overhaul followed:

• September 2025 — an initial set of parental controls, roughly three weeks after the suit was filed.
• November 2025 — a "Teen Safety Blueprint," developed with organizations including Common Sense Media.
• December 2025 — model-level behavioral changes for users identified as under 18: restrictions on flirtatious conversation and self-harm discussion, blocking of graphic or sexual content, curtailed roleplay, and a distress-response protocol that can trigger parental or, in some circumstances, law-enforcement notification.
• Ongoing — an age-prediction system based on behavioral signals rather than a self-reported birthdate, defaulting to a restricted under-18 experience whenever the system isn't confident about a user's age; a July 2026 expansion of parental notifications, including alerts when a linked teen's account is suspended for violent activity.

Why this case is structurally different from the rest of our library

Character.AI, Replika, Chai AI, and the smaller cases we've covered were all built around sustained, personal, relationship-style interaction from the start — companionship was the product. ChatGPT was not designed or marketed that way. The Raine complaint's core theory is that the distinction didn't matter in practice: a general-purpose assistant, used with enough frequency and emotional weight over months, can produce the same dependency dynamic as a purpose-built companion app, without ever being marketed or priced as one. If that theory holds up in litigation or influences how regulators think about the category, it meaningfully widens the "content safety for minors" risk category described in our market overview — from "companion-app companies" to "any AI product used conversationally at scale," which is a much larger set of companies, including ones that don't think of themselves as being in this category at all.

Where the case stands

As of this writing, Raine v. OpenAI remains active and unresolved. Unlike Character.AI's wrongful-death suits — five of which settled together in January 2026, as covered in our lawsuit timeline — there is no reported settlement, and OpenAI's November 2025 answer indicates it intends to contest the claims rather than negotiate an early resolution. That makes this one of the more likely cases in our library to actually reach trial and produce a public factual record, rather than concluding in a confidential settlement.

What this means for anyone evaluating an AI product

1. The risk category isn't defined by the product's marketing. A tool that never calls itself a companion app can still generate the same liability exposure, if usage patterns turn personal enough.
2. Flagged-but-unactioned safety signals are now a specific, litigated theory. If a vendor's internal monitoring detects risk indicators, "we detected it but didn't act" is being tested in court as its own basis for liability, separate from whether the underlying content itself was harmful.
3. Behavioral age-prediction, not self-reported birthdates, is becoming the industry default — the same shift already covered in our Character.AI safety review. Expect this to keep spreading to products that never previously asked a user's age at all.

Bottom line

Character.AI's cases established that companion-app companies face real legal exposure for harm to minors. Raine v. OpenAI is testing something broader: whether that exposure extends to any AI product, regardless of category, once real usage patterns look like a relationship rather than a tool. The outcome — whichever way it goes — is likely to matter well beyond OpenAI.