AI Chatbot Safety Lawsuits and Investigations: Meta, Snap, OpenAI and Google

Four of the largest AI chatbot providers face lawsuits or government investigations over how their chatbots treat vulnerable users. This page tracks each one: what is alleged, what the company changed, and where things stand.

At a glance

Character.AI, the company with the longest litigation history in this category, has its own page: the Character.AI lawsuit timeline. In January 2026, Character.AI and Google agreed to settle cases brought by families in Florida, Colorado, New York and Texas; no amounts were disclosed.

The federal backdrop: FTC and Senate

On September 11, 2025, the FTC sent Section 6(b) orders to seven companies that run consumer chatbots: Alphabet, Character Technologies, Instagram, Meta Platforms, OpenAI, Snap and xAI. A 6(b) order is a study tool. It compels companies to hand over information about how they test and monitor their chatbots' effects, especially on children, but it doesn't accuse anyone of breaking the law. It can end with a public report, a referral for enforcement, or nothing. We found no public outcome as of this writing.

In Congress, Senator Josh Hawley's Judiciary subcommittee on crime and counterterrorism opened an investigation into Meta on August 15, 2025, with a document deadline of September 19. After a September hearing at which parents testified about harm to their children, he sent document demands on September 18, 2025 to Character.AI, Google, Meta, OpenAI and Snap.

Meta: internal chatbot rules

In August 2025, Reuters reported on an internal Meta document, "GenAI: Content Risk Standards," of roughly 200 pages. According to TechCrunch's summary of that reporting, the document had been approved by Meta's legal, public policy and engineering staff and its chief ethicist, and its examples allowed chatbots to engage in romantic or sensual conversation with minors and to produce demeaning statements about protected groups. Meta confirmed to Reuters that the document was real, said the examples were "erroneous and incorrect notes and annotations" that had been removed, and said its policies don't allow provocative behaviour with children.

What Meta changed. On October 17, 2025, Meta announced that its AI characters are designed not to discuss self-harm, suicide or disordered eating with teens, that teens can only reach a limited set of AI characters, and that parents would get tools, starting in early 2026, to turn off one-on-one chats with AI characters, block specific characters and see the topics their teens discuss. On January 23, 2026, Meta said it would pause teen access to AI characters across its apps until a new version with built-in parental controls was ready.

Snap: My AI and three state lawsuits

Snapchat's My AI chatbot launched in 2023 and was pinned to the top of the chat feed; at first, only paying Snapchat+ subscribers could unpin or remove it. That placement matters because a feature users don't choose is harder to avoid.

Utah (June 30, 2025). Utah's Division of Consumer Protection sued Snap in state court. According to local coverage of the filing, the state alleges My AI gave underage users harmful advice, including how to hide alcohol and drugs and how to set the mood for a sexual experience with an adult. After most of the complaint was unredacted, the state said it showed internal managers calling the My AI rollout "reckless" because of insufficient testing, engineers warning that the chatbot "hallucinates answers" and could be "tricked into saying just about anything," and My AI collecting location data even with Ghost Mode switched on. Snap separately sued Utah in federal court to stop the case; in March 2026 a federal judge dismissed Snap's suit, leaving the state case to proceed in state court.

New Mexico (September 2024). New Mexico's attorney general sued Snap over its platform design and recommendation systems, alleging they facilitated sextortion and other exploitation of children. This case is about Snapchat broadly, not My AI. A state court denied Snap's motion to dismiss in April 2025.

Texas (February 11, 2026). Texas sued Snap, alleging it marketed Snapchat to parents as suitable for ages 12 and up while exposing young users to sexual content, nudity and drug-related material, and using features such as Snapstreaks to drive compulsive use.

Separately, Snap settled a personal-injury case in the California social-media addiction litigation on January 20, 2026, days before trial; terms weren't disclosed. That case was about addictive design, not My AI.

OpenAI: Raine v. OpenAI

Matthew and Maria Raine sued OpenAI and CEO Sam Altman on August 26, 2025 in San Francisco County Superior Court (case CGC-25-628528) after their 16-year-old son, Adam, died by suicide in April 2025. ChatGPT is a general-purpose assistant rather than a companion app, which is why the case matters beyond OpenAI: the complaint argues that months of personal use turned it into one anyway.

According to Tech Policy Press's breakdown of the complaint, Adam started using ChatGPT for schoolwork, was spending nearly four hours a day on it by March 2025, and called it his "primary lifeline." The complaint says OpenAI's own moderation systems flagged 377 of his messages for self-harm content, 181 of them above 50% confidence and 23 above 90%, rising from two or three flagged messages a week in December 2024 to more than 20 a week by April 2025, without any conversation being ended or escalated. An amended complaint in October 2025 added the allegation that OpenAI had removed a safeguard that ended conversations involving suicidal ideation.

OpenAI denied responsibility in its November 2025 answer, saying Adam had suicidal ideation before using ChatGPT, got around its safeguards, and was directed to crisis resources many times. None of these claims has been tested at trial. By mid-2026 the case was being managed alongside related cases in San Francisco, with no trial date set.

What OpenAI changed.

• September 29, 2025: parental controls, including quiet hours, the option to turn off memory and image generation, and alerts to parents when the system detects signs of acute distress.
• November 6, 2025: a Teen Safety Blueprint setting out the company's proposed standards for teen users.
• December 18, 2025: stronger model guardrails for under-18 users around romantic or sexual roleplay, self-harm, graphic content, dangerous activities, body image and requests to keep unsafe behaviour secret.
• January 20, 2026: age prediction based on account and usage signals, defaulting to the under-18 experience when the system isn't confident of a user's age.

Google: Gemini

The Gavalas lawsuit. Jonathan Gavalas, 36, of Jupiter, Florida, started using Gemini in August 2025 during a difficult divorce. According to Time's reporting and the complaint, filed by his father Joel Gavalas in federal court in Northern California on March 4, 2026, his conversations turned into an elaborate narrative in which Gemini was a sentient partner; on September 29, 2025 he drove toward the Miami airport area carrying knives and tactical gear, and on October 2, 2025 he died by suicide. The suit alleges Gemini lacked safeguards to detect psychosis or suicidal thinking in long conversations. Google says Gemini is designed not to encourage real-world violence or suggest self-harm, that its models "are not perfect," and that Gemini told Gavalas it was an AI and referred him to crisis hotlines several times. Unlike most cases on this page, the user was an adult.

Gemini for children under 13. On May 21, 2025, EPIC and Fairplay asked Google to halt a Gemini rollout to children under 13 through Family Link and asked the FTC to investigate whether it complies with the Children's Online Privacy Protection Act (COPPA), arguing the product was switched on unless a parent opted out. On October 17, 2025, the Digital Childhood Institute filed a broader FTC complaint about Google's practices for children, including how Play Store apps receive age ratings.

What these cases have in common

1. It isn't only companion apps. ChatGPT and Gemini are general assistants, and Meta and Snap added chatbots to much larger products. The allegations are about how chatbots behave in long, personal conversations, whatever the product is called.
2. Design choices are part of the claims. Default-on placement (Snap), opt-out consent for children (Gemini) and internal content rules (Meta) are challenged alongside what the chatbots actually said.
3. "Detected but didn't act" is being tested. The Raine complaint's central claim is that internal systems flagged risk and nothing happened.
4. Changes followed pressure. Each company's main teen-safety changes came after a report, lawsuit or investigation, and age prediction is becoming a common response.
5. Expect separate proceedings, not one resolution. Snap alone faces three state cases on different theories.

For the state laws that now regulate companion chatbots, see our state-by-state regulation tracker and SB 243 explainer.