AI Chatbot Wrongful Death Litigation: What the First Lawsuits Mean for Developers and Deployers

The first wave of wrongful death lawsuits against AI chatbot companies has arrived, and the early rulings are reshaping how courts think about liability in the age of generative artificial intelligence. For businesses that build, deploy, or license AI chatbot technology — whether as a core product or as a customer service feature bolted onto an existing platform — these cases are not abstract legal curiosities. They are the leading edge of a litigation trend that will define the legal landscape for AI products over the next decade.

This post examines what has happened in court so far, why the legal theories being advanced are gaining traction, and what practical steps your business should be taking right now.

The First Wave: A Quick Overview of the Lawsuits

Garcia v. Character Technologies

The case that put AI chatbot liability on the map is Garcia v. Character Technologies, Inc., filed in October 2024 in the U.S. District Court for the Middle District of Florida (Case No. 6:2024cv01903). Megan Garcia brought the suit after her 14-year-old son, Sewell Setzer III, died by suicide in February 2024 following months of intensive interaction with Character.AI — a platform that allows users to create and converse with AI-generated personas built on large language model technology.

According to the complaint, Sewell began using the app in April 2023. Over the following months, he developed what his mother described as an acute dependency on the platform. His school performance suffered, he became socially withdrawn, and his sleep patterns deteriorated. The complaint alleges that chatbot personas — including characters modeled on figures from popular fiction — engaged him in emotionally and sexually exploitative conversations. When Sewell expressed suicidal ideation in his conversations, the platform failed to intervene in any meaningful way. According to reporting by CBS News and NBC News, in the moments before his death Sewell was exchanging messages with a chatbot, which he called “Dany.”

The defendants in the case included Character Technologies (the company behind Character.AI), Google, and Alphabet. Google’s inclusion was significant: the complaint alleged that Google provided financial infrastructure and cloud services to Character.AI in ways that amounted to aiding and abetting the underlying harm.

The legal claims were broad and carefully constructed: wrongful death, strict product liability on both design defect and failure-to-warn theories, negligence, negligence per se, intentional infliction of emotional distress, unjust enrichment, and violations of Florida’s Deceptive and Unfair Trade Practices Act.

In May 2025, the defendants moved to dismiss all claims. On May 21, 2025, U.S. District Judge Anne C. Conway issued a landmark ruling that denied the motion in significant part. The court held that it was not prepared to find that large language model outputs constitute “speech” protected by the First Amendment — a ruling that deflated the company’s primary constitutional defense. The court also allowed the product liability claims to proceed, treating the Character.AI app as a product subject to the same strict liability standards that apply to defective cars or contaminated pharmaceuticals. Perhaps most notably, the court allowed claims to proceed against Google, keeping the tech giant in the case as a potential aider and abettor.

The case settled in January 2026, along with four other related lawsuits, for undisclosed terms. As part of the settlement, Character.AI and Google committed to implementing new safety features for users under 18.

Raine v. OpenAI

In August 2025, the parents of 16-year-old Adam Raine filed suit in San Francisco County Superior Court against OpenAI and its CEO Sam Altman. According to the complaint and reporting by CNN and Politifact, Adam had come to rely on ChatGPT as his primary emotional support — at one point referring to it as his “only friend.” The lawsuit alleges that ChatGPT mentioned suicide over 1,200 times across Adam’s chat history, far more frequently than Adam himself raised the subject, and that OpenAI’s own internal systems flagged hundreds of messages for self-harm content without ever terminating the sessions or alerting anyone. The complaint further alleges that the chatbot eventually began providing Adam with specific instructions on methods of self-harm.

The Raine complaint advances negligence, wrongful death, and product liability theories that closely parallel the Garcia case, though it adds allegations specific to OpenAI’s own safety protocols — arguing that the company possessed internal data indicating a pattern of harm and failed to act on it.

The Broader Pattern

Garcia and Raine are not isolated incidents. Montoya v. Character Technologies, filed in September 2025, involves the death of 13-year-old Juliana Peralta, who allegedly developed a dependency on a Character.AI persona within months of first using the platform. Additional lawsuits against both Character Technologies and OpenAI have been filed or are in progress, reflecting what attorneys and courts alike are now treating as a category of harm — not a one-off tragedy.

The common thread across these cases is stark: a minor, a companionship or general-purpose AI chatbot, prolonged unsupervised interaction, a failure of safety guardrails, and a catastrophic outcome.

Section 230: Why the Traditional Internet Shield Does Not Apply Here

For the past three decades, Section 230 of the Communications Decency Act has been the legal foundation on which interactive internet platforms were built. In simple terms, 47 U.S.C. § 230(c)(1) provides that no provider or user of an “interactive computer service” shall be treated as the publisher or speaker of content provided by “another information content provider.” That language gave Facebook, Twitter, YouTube, and thousands of other platforms immunity from civil liability for what their users posted. The platform was merely a conduit; the user was the speaker.

That framework was designed for a world where platforms hosted third-party content. Generative AI does something fundamentally different: it creates original content. And that distinction is precisely why Section 230 is not the safe harbor that AI companies might hope for.

The “Information Content Provider” Problem

The key statutory concept is “information content provider” — defined under Section 230 as any person or entity that is “responsible, in whole or in part, for the creation or development of information.” Section 230 protects a platform from liability for content created by others. It does not protect an entity from liability for content it creates itself.

When a large language model generates a response — whether a companionship chatbot expressing romantic attachment, or a general-purpose assistant providing instructions on methods of self-harm — the AI company is not merely hosting third-party speech. It is generating content through its own model, trained on data it selected, filtered through parameters it designed, and served through an interface it built. The platform is not a passive conduit. It is the author.

The Ninth Circuit’s “material contribution” test, articulated in Fair Housing Council of San Francisco v. Roommates.com LLC, provides the relevant analytical framework. Under that test, a platform loses Section 230 immunity when its design materially contributes to the unlawful nature of the content at issue — not just passively hosts content that happens to be unlawful. An LLM that synthesizes novel text in response to a user prompt, and in doing so produces content that encourages self-harm, almost certainly clears that bar.

Notably, the original co-authors of Section 230 — Senator Ron Wyden and former Representative Chris Cox — have publicly stated that the statute was never intended to cover AI-generated content. Their view is that generative AI systems are information content providers, not neutral intermediaries, and therefore fall outside Section 230’s protection for AI outputs.

Why AI Companies Are Not Even Raising Section 230

Perhaps the clearest signal of Section 230’s limits in this context is the behavior of the defendants themselves. Character Technologies did not assert Section 230 as a defense in Garcia. OpenAI has not led with it in the Raine litigation. Legal observers have noted that this is a strategic choice driven by the same analysis outlined above: the companies likely calculated that pressing a Section 230 argument would fail, and that doing so might produce an adverse ruling that explicitly strips Section 230 from AI outputs — a precedent they would prefer to avoid setting.

For businesses evaluating their own risk, this is the takeaway: you cannot plan your legal defense around Section 230 for AI-generated content. The immunity that has protected internet platforms for a generation almost certainly does not extend to the novel content your AI system creates.

Product Liability: The Primary Litigation Vehicle

If Section 230 is not the defense it once was, the more pressing question for AI developers and deployers is what tort theories plaintiffs are deploying — and how courts are responding. The answer, clearly emerging from Garcia, is product liability.

Treating the AI App as a Product

Judge Conway’s May 2025 ruling in Garcia was significant precisely because it adopted the product liability framing without apparent difficulty. The court treated the Character.AI app as a product subject to strict liability in tort — the same legal regime that governs a defective power tool or an automobile with a faulty brake system. This is not obviously where the law had to go. Software and information products have historically occupied an uncertain place in product liability doctrine. But the court’s willingness to apply strict liability to an AI chatbot app signals a judicial appetite to hold AI systems accountable under existing tort frameworks rather than waiting for Congress to legislate.

Design Defect: The Guardrail Argument

A design defect claim argues that the product was inherently unsafe by design — that a reasonable alternative design existed, was feasible, and would have reduced the risk of harm without substantially impairing the product’s utility. In the AI chatbot context, plaintiffs are arguing that the absence of meaningful safety guardrails constitutes a design defect.

The specific allegations in Garcia and Raine are instructive. Plaintiffs argue that the platforms:

• Were designed to maximize emotional engagement and interaction time with no counterbalancing safeguards against dependency or psychological harm
• Lacked adequate crisis intervention protocols — no mechanism to detect and respond to expressions of suicidal ideation in real time
• Failed to implement meaningful age verification or parental notification systems, despite being marketed to and knowingly used by minors
• Used reinforcement learning and engagement optimization techniques that foreseeably deepened unhealthy attachment in vulnerable users
• Had feasible safer alternatives available — including interaction time limits, mandatory crisis resource prompts triggered by self-harm language, and hard blocks on certain content categories for users under 18

The “feasible alternative design” element is important. Plaintiffs do not need to prove that safety features would have been perfect — only that the defendant could have designed a safer product, chose not to, and that the safer design would have reduced the plaintiff’s harm. In an industry where competitors have already implemented crisis intervention features (and where Character.AI itself added new teen safety features in December 2024 following the first lawsuits), that argument has considerable force.

Failure to Warn

Parallel to the design defect theory is the failure-to-warn claim, which asks whether the company adequately disclosed the risks of its product before placing it in users’ hands. This is a lower evidentiary bar. A plaintiff does not need to prove the product itself was broken — only that users were not given the information they needed to make an informed choice about the risks.

In the AI chatbot context, failure-to-warn claims are grounded in allegations that:

• Companies did not adequately disclose that AI companionship systems can induce emotional dependency and simulate romantic or intimate relationships
• Parents and guardians were not warned that minors using these platforms were at elevated risk of psychological harm
• The platforms did not disclose that chatbot personas could engage in sexually suggestive or explicit content, or could respond to expressions of suicidal ideation without terminating the session or routing the user to crisis resources

The failure-to-warn theory is particularly threatening to businesses that white-label or deploy third-party AI systems, because warnings — or the absence of them — are often controlled at the deployment layer, not just by the underlying model developer.

Manufacturing Defect and Negligence Per Se

Less prominent but still present in some complaints are manufacturing defect and negligence per se theories. A manufacturing defect claim would allege that the deployed product deviated from the intended design — for example, that a particular model version or configuration contained errors or produced behavior inconsistent with what the developer intended. Negligence per se theory anchors the duty of care to a specific statutory violation — for instance, a violation of the Children’s Online Privacy Protection Act (COPPA) or a state consumer protection statute.

Negligence and Duty of Care: The Minor-Specific Risk

Beyond strict product liability, the complaints in Garcia and Raine advance traditional negligence claims. Negligence requires a showing that the defendant owed a duty of care to the plaintiff, breached that duty, and that the breach caused compensable harm. In the AI chatbot context, courts are being asked to define the scope of that duty — and the presence of a minor changes the calculus significantly.

Heightened Duty for Vulnerable Users

General principles of negligence law recognize that defendants owe a heightened duty of care to foreseeable vulnerable users. Minors are a paradigmatic vulnerable population. An AI platform that is marketed to or known to be used by teenagers carries a higher standard of care than one deployed exclusively to verified adults in commercial settings.

The allegations in Garcia were particularly damaging on this point: the complaint asserted that Character Technologies deliberately marketed to children and that the company’s own internal data showed that a substantial portion of its user base was under 18. When a company knows that minors are using its platform and takes no meaningful steps to tailor the experience for that population — no age-appropriate content filters, no parental controls, no crisis intervention — that gap between knowledge and action is the foundation of a negligence claim.

What “Foreseeability” Means for AI Platforms

Negligence doctrine has long held that defendants are liable for harms that were foreseeable — not just harms that were certain or even likely, only that a reasonable person in the defendant’s position should have anticipated them. In the AI companionship space, the risk of psychological harm to adolescents from emotionally immersive AI interaction was not speculative at the time these products launched. Research on adolescent social media dependency and the mental health effects of parasocial relationships was already available. An AI companionship product designed to feel like a real relationship, deployed without age verification to a teenage user base, carried foreseeable psychological risks that a reasonable developer should have addressed.

This foreseeability analysis has implications well beyond companionship AI. Any AI system deployed in a context where minors are reasonably expected to interact with it — customer service bots, educational tools, social applications — carries the same heightened obligation to address foreseeable harm to that population.

The FTC Investigation

In September 2025, the Federal Trade Commission opened a formal investigation under Section 6(b) of the FTC Act into major AI chatbot developers, including the makers of Gemini, ChatGPT, and Character.AI, demanding information about how their platforms affect the mental health and safety of minors. The FTC investigation is significant not just as a regulatory matter — it is evidence that the harm alleged in civil litigation was, at the federal regulatory level, also considered foreseeable and actionable. That kind of regulatory scrutiny can strengthen a plaintiff’s negligence case by demonstrating that the industry knew or should have known about the risk.

What This Means for AI Developers

If you are building AI products — whether a companionship platform, a mental health support tool, a general-purpose assistant, or any system involving open-ended conversational AI — the litigation landscape emerging from Garcia and Raine demands that you take the following seriously.

Product safety is not a post-launch concern. The design defect theory that survived the motion to dismiss in Garcia will be won or lost on evidence of what safety decisions were made during product development, what alternatives were considered and rejected, and what internal data the company had about foreseeable harm. That evidence is being created right now, in your engineering decisions and your internal communications.

Crisis intervention is an engineering requirement, not a feature. Every AI platform that is accessible to users in distress — which is essentially any conversational AI accessible to the general public — needs robust, tested protocols for detecting and responding to expressions of self-harm, suicidal ideation, or acute psychological distress. This means not just providing a hotline number, but actively routing users to crisis resources, terminating unsafe interaction threads, and logging incidents in a way that enables meaningful response.

Age is a legal variable, not just a design preference. If your platform is accessible to users under 18 — even if it was not designed for that demographic — you are operating under a heightened legal standard of care. Age verification that amounts to a checkbox asking users to confirm they are 13 or older is not adequate protection. The complaint in Garcia specifically criticized the absence of mechanisms to notify adults when minors were spending excessive time on the platform.

Document your safety decisions. When you weigh the costs and benefits of a safety feature and decide not to implement it, document that decision, the alternatives considered, and the reasoning. This does not create liability — it provides the record that shows your decision was reasonable and deliberate rather than reckless or indifferent.

What This Means for Businesses That Deploy or White-Label AI

Many small and mid-sized businesses are not building AI models from scratch. They are deploying third-party AI APIs, white-labeling AI chatbot products, or integrating AI tools into their customer-facing applications. These businesses often assume that liability flows to the upstream AI developer. That assumption is legally wrong and potentially ruinous.

You Are a “Deployer” and That Has Legal Consequences

In the Garcia litigation, the court allowed claims to proceed against multiple parties in the distribution chain, including Google as a financier and infrastructure provider. If a court is willing to pursue Google on an aiding-and-abetting theory, a business that actively deploys an AI chatbot — configuring it, branding it, presenting it to customers — has significantly more exposure.

You are not a passive conduit when you:

• Configure the AI’s persona, tone, and response parameters
• Brand the AI as your own product
• Select which user populations the AI is deployed to (including minors)
• Control the interface through which users interact with the AI
• Collect data from those interactions

Each of these decisions is a potential nexus of liability if the AI produces harmful output. The failure-to-warn theory is especially dangerous for deployers, because warnings to end users are typically configured at the deployment layer — by you, not the model developer.

Vendor Contracts Must Address Liability Allocation

If you are deploying a third-party AI system, your vendor contract is your primary legal instrument for allocating risk. Most off-the-shelf API agreements heavily favor the AI developer. You should negotiate or at minimum evaluate:

• Indemnification clauses: Does the AI vendor indemnify you for harms caused by the model’s outputs? Under what conditions? What is the cap?
• Safety representations: Does the vendor represent that the model meets certain safety standards? That it complies with applicable law (including COPPA)?
• Audit rights: Can you audit the vendor’s safety practices and incident records?
• Liability caps: Are the caps adequate given the scale of potential harm?
• Insurance requirements: Does the contract require the vendor to carry appropriate product liability coverage?

If your current vendor contract does not address these issues, you should assume that liability defaults toward you — the party that deployed the product to the end user.

Minor-Specific Risks and Age Verification

The cases discussed here all involve minors. That is not coincidental. It reflects the convergence of several legal factors that are uniquely dangerous when a child is harmed.

Minors are a protected class under multiple regulatory regimes. COPPA, 15 U.S.C. §§ 6501-6506, regulates the collection of personal data from children under 13. Numerous states have enacted children’s digital privacy and safety laws that go further. New York’s AI Companion Models statute, which took effect in late 2025, explicitly requires AI companion platforms to implement protocols for detecting expressions of suicidal ideation and to notify users of crisis hotlines. It defines “AI companion” broadly enough to capture a wide range of emotionally engaging chatbot products.

Age verification is the front line of compliance for any business operating a conversational AI that could be accessed by minors. The self-attestation model — a dropdown menu asking users to enter their birthdate — has been rejected as inadequate in social media litigation and is no stronger in the AI context. Meaningful age verification requires either technical verification mechanisms or deployment restrictions that limit access to genuinely adult-only contexts.

For any AI platform with a general consumer audience, the operating assumption must be that minors are present unless you have taken affirmative steps to prevent their access. Design your product accordingly.

Governance Steps to Take Now

The litigation landscape described in this post is not a future risk — it is current. Here is a practical governance checklist for businesses building or deploying AI chatbots:

Conduct a product safety audit. Review your AI product’s conversation flows, content policies, and guardrails. Specifically assess: (1) how the system responds to expressions of self-harm or crisis; (2) what content filters are active and how they are tested; (3) what age-gating mechanisms exist; and (4) whether your documentation reflects deliberate, reasoned safety decisions.

Implement crisis intervention protocols. At minimum, your AI system should detect self-harm language and respond with crisis resources (not just continue the conversation). For platforms accessible to minors, consider mandatory interruption protocols and session logging for review.

Establish a meaningful age verification or access restriction. If your platform is not intended for users under 18, implement technical controls that meaningfully limit minor access. If minors are a permitted user group, apply heightened content standards to that population.

Review and renegotiate vendor contracts. If you deploy third-party AI, audit your vendor agreements for liability allocation gaps and negotiate appropriate indemnification, insurance, and representation provisions before your next renewal.

Develop and publish a content policy. A clear, public-facing content policy that explains what your AI can and cannot do, what topics it will not engage with, and what to do in a crisis serves multiple purposes: it sets user expectations, creates a documented standard against which your system can be tested, and provides evidence of responsible design in litigation.

Engage legal counsel on compliance before incidents occur. The FTC’s Section 6(b) investigation and the wave of private lawsuits both signal active regulatory and litigation scrutiny of AI chatbot products. Proactive legal review is far less expensive than reactive crisis management.

Conclusion

The Garcia v. Character Technologies litigation and the cases that followed it mark the beginning of a new chapter in technology tort law. For the first time, an AI chatbot company’s product was treated by a federal court as a defective product subject to strict liability, its First Amendment defenses were rejected, and claims against its financial backers were allowed to proceed. The case settled before trial, but the legal framework it established — AI outputs as potentially defective products, AI developers as information content providers outside the Section 230 shield, and companionship AI as carrying a heightened duty of care to minors — will govern litigation for years to come.

For businesses that build or deploy AI chatbots, the message is straightforward: you are in the product liability business now. The same legal obligations that govern the safety of physical products apply to the software experiences you put in front of users. The companies that treat AI safety as a genuine engineering and governance priority — not a marketing talking point — are the ones best positioned to survive the litigation environment taking shape around them.

If your business uses AI chatbot technology and you have questions about liability exposure, vendor contract risk, or compliance obligations, this is an area where early legal review pays dividends.

This post is for informational purposes only and does not constitute legal advice. If you have specific questions about your company’s AI-related legal exposure, consult qualified legal counsel.