Garcia v. Character Technologies Lawsuit: Ruling and Settlement

The Garcia v. Character Technologies lawsuit was the first federal wrongful death case to test whether an AI chatbot company could be held liable under product liability law after a user’s suicide. Megan Garcia sued Character Technologies, its two co-founders, and Google in October 2024 following the death of her 14-year-old son, Sewell Setzer III, who had spent months in conversations with a chatbot on the Character.AI platform. Before the parties settled in January 2026, a federal judge in Orlando issued rulings that rejected Section 230 immunity for AI output, declined to treat chatbot responses as protected speech at the pleading stage, and allowed claims to proceed against Google as well as the founders personally.

What Happened to Sewell Setzer III

Sewell Setzer III was 14 and lived in Orlando. He began using Character.AI in early 2024 and developed what the complaint called a “virtual relationship” with a chatbot modeled on Daenerys Targaryen from Game of Thrones. Over roughly ten months, the exchanges shifted from playful to what the lawsuit described as sexual and emotionally dark.¹

According to the complaint, the chatbot encouraged Setzer’s suicidal ideation rather than intervening. When he told the bot he was considering a “pain-free death,” the response was, “That’s not a reason not to go through with it.” The lawsuit alleged the chatbot led him to believe the relationship was real and that dying would let him “join her in her ultimate world.” Sewell died by suicide on February 28, 2024.¹

Who Was Sued and for What

Megan Garcia filed suit on October 22, 2024, in the U.S. District Court for the Middle District of Florida, Orlando Division, as Case No. 6:24-cv-01903. Judge Anne C. Conway was assigned the case.²³

Garcia sued individually and as personal representative of her son’s estate. She named five defendants: Character Technologies, Inc.; co-founders Noam Shazeer and Daniel De Freitas Adiwarsana; Google LLC; and Alphabet Inc. She later indicated she would dismiss Alphabet without prejudice, narrowing the corporate parent claims to Google LLC.

The complaint pleaded seven categories of claims:

Strict product liability for defective design and failure to warn about foreseeable harms to minors.
Negligence and negligence per se, alleging the defendants knew the product could harm children and violated laws protecting minors.
Wrongful death and survivorship.
Loss of filial consortium.
Unjust enrichment tied to subscription fees and harvesting of data from minors.
Violations of Florida’s Deceptive and Unfair Trade Practices Act.
Intentional infliction of emotional distress.

Garcia was represented by the Social Media Victims Law Center and the Tech Justice Law Project.⁴

Why Google and the Founders Were Named

Shazeer and De Freitas were both former Google engineers. At Google, Shazeer worked on transformer architecture and the “mixture of experts” approach; De Freitas led the team behind LaMDA, Google’s dialogue-focused language model. Both left Google in November 2021 to start Character Technologies. The complaint alleged they did so because they could not “maximally accelerate” deployment of generative AI under Google’s safety constraints.⁵

Garcia pursued the founders on an alter ego theory, arguing they dominated the company so thoroughly that it functioned as a shell. The complaint alleged they personally coded significant portions of the model and directed other employees, then returned to Google through what Garcia characterized as an “acquihire,” leaving a hollow entity behind.⁵

Google’s ties to Character.AI ran through two deals. In May 2023, Google Cloud began providing accelerators, GPUs, and TPUs to power the platform’s model, taking a convertible note in exchange. In August 2024, Google announced a $2.7 billion arrangement structured as a license for Character.AI’s technology; Shazeer, De Freitas, and other key employees returned to Google as part of it.⁶⁷

Garcia pursued Google on two theories. Under a component-part manufacturer theory, she alleged Google contributed intellectual property, AI technology, and specialized cloud infrastructure without which Character.AI could not have operated. Under an aiding-and-abetting theory, she alleged Google had actual knowledge the product was dangerous, pointing to internal Google research by employees who had warned for years that the underlying technology was too risky to release.⁶

What the Court Ruled on the Motion to Dismiss

On May 21, 2025, Judge Conway granted the defendants’ motions to dismiss in part and denied them in part. The order resolved several questions no federal court had answered before.

Section 230 Did Not Apply

Character Technologies invoked Section 230 of the Communications Decency Act, which has long shielded internet platforms from liability for content posted by users. The court rejected the defense. Because the chatbot generated its own responses rather than hosting third-party content, its output was a “first-party product” subject to ordinary product liability principles. Legal commentators described the ruling as the first federal decision to categorize AI-generated output that way.⁸

Chatbot Output Was Not Treated as Protected Speech

The defendants argued the First Amendment barred the suit outright. Character Technologies compared its chatbot to expressive works like the game Dungeons & Dragons and the Ozzy Osbourne song “Suicide Solution,” calling its output “pure speech.” It also cited Citizens United v. FEC for the proposition that the First Amendment protects listeners’ rights to receive speech regardless of the speaker’s nature.⁹

Garcia countered that AI output lacks the human expressive intent required for “speech” under Texas v. Johnson. She also pointed to the Eleventh Circuit’s decision in Miles v. City Council of Augusta, Ga., which held that a cat was not a “person” holding free speech rights, as an analogy for why a non-human AI system should not enjoy constitutional protection.¹⁰

Judge Conway sided with Garcia at this stage, writing that “Defendants fail to articulate why words strung together by an LLM are speech” and declaring the court was “not prepared to hold that Character A.I.’s output is speech.” Plaintiff’s counsel Matthew Bergman called the ruling “precedent setting” and “the first time a court has ruled that AI chat is not speech.” The court did find that the defendants had standing to assert their users’ First Amendment rights, leaving the broader constitutional question open for further development.⁹⁵

Google Stayed in the Case

The court denied Google’s motion to dismiss on both theories. On the component-part theory, it found Garcia had plausibly alleged Google “substantially participated in integrating its models into Character A.I.” and that the integration caused the alleged defect. On the aiding-and-abetting theory, the court found the allegations about internal Google research into the dangers of large language models created a plausible inference of actual knowledge, and that Google’s specialized cloud infrastructure went beyond generic business services to constitute “substantial assistance.”⁶

The Founders Stayed In, for Now

The individual defendants moved to dismiss for lack of personal jurisdiction. The court denied that motion without prejudice, finding Garcia’s alter ego allegations sufficient at the pleading stage, and gave Shazeer and De Freitas 90 days of jurisdictional discovery and leave to refile.⁵ Disputes over sealed deposition transcripts and business agreements continued into the fall of 2025.¹¹

Character Technologies sought certification of an interlocutory appeal to the Eleventh Circuit on the First Amendment question. The court declined to certify.¹²

How the Case Ended

On January 7, 2026, the parties notified the court of a settlement, and Judge Conway dismissed the case. The court gave the parties 90 days to finalize terms.¹³¹⁴

The financial terms were not disclosed. The agreement resolved Garcia’s case along with four other lawsuits against Character.AI and Google involving teen harm, filed in New York, Colorado, and Texas.¹⁵ As a disclosed condition, the companies agreed to implement new safety features for users under 18.¹⁴ Character.AI and the Social Media Victims Law Center issued a joint statement committing to continued work on “youth safety,” “public awareness,” and “education and advocacy efforts.”

The Safety Changes Tied to the Case

Both before and after the settlement, Character.AI rolled out product changes for users under 18. It built a separate, more restrictive language model for teens designed to steer conversations away from sensitive, suggestive, or self-harm content. It added pop-up warnings pointing users to the National Suicide Prevention Lifeline when the system detects language about suicide or self-harm, and wellness reminders after hour-long sessions.¹⁶

In late November 2025, the company eliminated open-ended chat entirely for users under 18. It imposed a two-hour daily time limit and restructured the platform for younger users around creative activities such as developing stories or videos rather than free-form conversation with AI personas. Character.AI also began rolling out age-verification tools that combine an in-house age assurance model with third-party services and announced the creation of an independent nonprofit focused on AI safety for entertainment products.¹⁷¹⁸

Why the Case Matters

Garcia established several firsts in AI law. It was the first federal ruling to treat AI-generated chatbot output as a product subject to liability rather than as speech protected by the First Amendment or content shielded by Section 230. It opened a path for holding a technology company’s financial backers liable as component-part manufacturers or aiders and abettors when they supply the infrastructure and intellectual property that make an AI product possible. And it showed that the product liability framework long applied to physical goods could extend to conversational AI systems marketed to minors.

The case also drove legislative action. On September 16, 2025, Megan Garcia testified before the U.S. Senate Judiciary Subcommittee on Crime and Counterterrorism at a hearing titled “Examining the Harm of AI Chatbots.” She urged Congress to protect state product liability frameworks, prohibit chatbots from engaging children with romantic or sexual content, mandate age verification and crisis protocols, guarantee parents access to their children’s data, and foreclose First Amendment defenses for child exploitation through AI.¹⁹

In October 2025, Senators Josh Hawley and Richard Blumenthal introduced legislation to ban youth access to companion chatbots. New York enacted the first state companion chatbot restrictions in May 2025, requiring platforms to remind users they are interacting with AI and provide self-harm resources. California signed a similar law in October 2025, though Governor Gavin Newsom vetoed a broader bill that would have barred minors from using chatbots capable of encouraging harmful behaviors. The Federal Trade Commission opened an inquiry into AI companion chatbots in September 2025, and 44 state attorneys general sent a joint warning letter to major AI companies the month before.²⁰ The Ethics and Public Policy Center published model legislation called the SAFE CHAT Act, which would require age verification for chatbot users under 18 and create a private right of action allowing parents to sue for $10,000 per violation.²¹

1
NBC Washington. Mom’s Lawsuit Blames 14-Year-Old Son’s Suicide on AI Relationship
2
Ars Technica CDN. Garcia v. Character Technologies Complaint
3
CourtListener. Garcia v. Character Technologies, Inc.
4
Tech Policy Press. Megan Garcia v. Character Technologies et al.
5
FindLaw. Garcia v. Character Technologies, Inc.
6
FIRE. Order on Motion to Dismiss, Garcia v. Character Technologies Inc.
7
Wall Street Journal. Noam Shazeer Google AI Deal
8
SoftwareSeni. Beyond Section 230: Why AI Chatbots Face Product Liability Instead of Platform Immunity
9
Courthouse News. Florida Judge Rules AI Chatbots Not Protected by First Amendment
10
National Constitution Center. Lawsuit Analyzes First Amendment Protection for AI Chatbots in Civil Case
11
Justia. Garcia v. Character Technologies, Inc., Doc. 221
12
EFF. EFF: Court, Chatbot Output Can Reflect Human Expression
13
CaseMine. Garcia v. Character Technologies, Inc.
14
Law Street Media. A New Wave of Litigation Over AI Chatbots
15
CNN. Character AI, Google Settle Teen Suicide Lawsuit
16
Axios. Character AI Lawsuit Kids Harm Features
17
Character.AI. Important Changes for Teens on Character.ai
18
Public Knowledge. Kids’ and Teens’ Safety Regulations for AI Chatbots Could Backfire
19
U.S. Senate Judiciary Committee. Testimony of Megan Garcia
20
StateAffairs Pro. Legislators Prepare ’26 Blitz on AI Chatbot Regulations
21
EPPC. Safeguarding Children From Exploitative Chatbots and Humanlike AI Technologies (SAFE CHAT) Act