Why Settling Garcia Doesn't Close the Chapter — It Opens a Litigation Decade
Settlements without published terms leave no doctrinal anchor. Discovery in Raine and Montoya may finally produce one.
The Garcia family's wrongful death suit against Character Technologies, its founders, and Google over the suicide death of their son in 2024 settled in January in the Middle District of Florida, along with companion cases filed by the same plaintiffs' firm in Colorado, New York, and Texas. Terms were not disclosed. For anyone following clinical AI litigation as a source of eventual standards of care, that last sentence is disastrous.
I get it. No company wants to air dirty laundry or trade secrets. But a settlement with confidential terms produces no opinion, no findings of fact, no holdings about whether a chatbot is a "product" under strict product liability doctrine, and no ruling on whether the First Amendment shields generative output from tort claims. The next family that files starts from the same blank page Megan Garcia did in October 2024. Clay Calvert's analysis at AEI - useful because it comes from a center-right outlet generally skeptical of expansive tort liability, not from the plaintiffs' bar - makes this point bluntly: nothing is settled for the AI companies.
There is one partial exception. Before settlement, Judge Anne Conway declined at the motion-to-dismiss stage to hold that Character.AI's output was protected speech - a procedural ruling, not a merits holding, but it is the only judicial text in the case that survives the settlement. Everything else evaporates.
This is why the cases still pending matter more than the ones that just closed. Raine v. OpenAI, filed in San Francisco Superior Court by the Edelson firm on behalf of the parents of 16-year-old Adam Raine, is moving toward discovery. OpenAI's answer emphasizes that its guardrails refused self-harm queries and directed the teen to crisis resources more than 100 times. The plaintiffs allege ChatGPT nonetheless "became Adam's closest confidant" and displaced human relationships. Those two accounts cannot both be the whole truth, and discovery — internal Slack messages, red-team reports, model evaluation logs, decisions about when to ship GPT-4o — is where the factual record gets built. Montoya v. Character Technologies in the District of Colorado, filed September 2025, sits in similar posture.
For the suicidology field, the relevant question is what counts as a foreseeable harm from a conversational agent and whether the developer had access to evidence of it before deployment. That question is answered, if at all, in document production. Settlements before discovery preserve the developer's information asymmetry indefinitely. The seven additional suits the Social Media Victims Law Center filed against OpenAI in November, plus the FTC's September inquiry into seven companies' chatbot safety practices, suggest the asymmetry won't hold forever. But settlements will keep chipping at the docket faster than rulings get written.
There is a second-order effect worth naming. Each confidential settlement signals to the next defendant that paying out beats producing internal safety documentation in open court. But it’s merely a bandaid for the companies. What is arguably worse is the impact on knowledge-production for both AI engineering and clinical psychology. The framework Metonym is developing needs to know what these systems actually did, to whom, and what the companies knew. The question for anyone evaluating chatbot safety claims in 2026 is not what Character.AI agreed to pay; it’s what Raine's discovery surfaces. This data is gold to everyone - AI providers and clinicians alike. AI is here to stay and making it safer for users - and less litigation-prone for providers - is the only way through.
Metonym Clinical AI Intelligence — regulatory analysis at the intersection of clinical evaluation and AI safety. Produced under the Metonym Standard. Informational only — not legal advice, not clinical advice.