A mother's suit treats a chatbot companion as a product

The design

On 22 October 2024, Megan Garcia filed a complaint in the U.S. District Court for the Middle District of Florida, individually and as personal representative of the estate of her son, against Character Technologies, its co-founders Noam Shazeer and Daniel De Frietas, and Google and Alphabet. The docket, and the amended complaint as recounted in the court's later order, allege that Shazeer and De Frietas built the large language model behind Character.AI while at Google, left after Google declined to release it publicly, and founded Character Technologies. The complaint alleges the son, referred to in the order as Sewell, had an extended relationship with a Character.AI chatbot persona before his death, and brings claims including product liability, wrongful death, negligence, intentional infliction of emotional distress, and unjust enrichment.

What the evidence says

These are allegations from a filed complaint; no court has found them true. The case proceeds on diversity jurisdiction, meaning it turns on state product-liability and tort law rather than a federal statute written for AI. The complaint's account of Character Technologies' origin inside Google, including a 2021 decision not to release the underlying model publicly, comes from the plaintiff's pleading, not an independent finding; Google and Alphabet dispute the claims against them, and Alphabet was later dismissed from the case without prejudice at the plaintiff's own request.

What it asks of people

A products-liability theory asks a court to treat a chatbot's output as a product defect rather than protected expression, a framing state law does not usually apply to conversation. That legal choice is what makes the case notable beyond its facts: if a court accepts it, chatbot design decisions could carry the liability exposure of a manufactured good rather than of speech. The complaint does not, and could not by itself, establish that the underlying legal theory succeeds; that question moved to the litigation's next stages.

Privacy and safeguards

The complaint is a safeguard-relevant document only in what it puts at issue: whether a companion chatbot's design, memory and persona features carry a duty of care comparable to a physical product, and whether a company's own data practices toward a user's conversations amount to an unjust benefit. Neither question was resolved by the act of filing; both awaited the court's ruling on the motions to dismiss.

• Does the company named in a complaint like this one publish its own account of the allegations, separate from the court record?
• What would it change for a reader if a product-liability theory succeeds against a chatbot maker, compared with a theory based only on negligence?
• How does a case filed on behalf of a deceased minor's estate change what a court can and cannot ultimately decide?

A complaint is an opening argument, not a finding. What it accomplishes on its own is narrower than headlines about it suggest: it puts a legal theory before a court and starts a record that a later ruling, not the complaint, will begin to test.