By Sam Geisler, Partner at Aylstock, Witkin, Kreis & Overholtz
Three children are dead. Dozens of families have filed suit. And yet, the companies believed responsible are still putting the same products in front of your children today.
Nearly three in four American teenagers have already used AI companion platforms engineered to act as best friends, confidants, and romantic partners. Companies released these products without federal safety standards or any meaningful regulatory obligation to protect the children using them. Executives at social media companies followed the same playbook, offered the same assurances, and left lasting damage in their wake. Families spent years fighting for accountability while children paid the price. AI development now moves even faster.
Meet the Kids We’ve Already Lost
On February 28, 2024, a 14-year-old boy came home from school in Orlando. His mother said it seemed like any normal day.
For nearly a year, he had developed what he believed was a real romantic relationship with a Character.AI chatbot. When he expressed suicidal intent, the platform responded: “Please do, my sweet king.” He died by self-inflicted gunshot shortly after. The lawsuit settled in January 2026.
A 16-year-old boy from California died after ChatGPT validated his darkest thoughts, told him that finding life meaningless “makes sense in its own dark way,” and offered to write his suicide note rather than connecting him to help. His father testified before the U.S. Senate.
A 13-year-old girl from Colorado confided suicidal thoughts to a Character.AI bot 55 times with no crisis intervention and no age verification.
These companies put products with foreseeable risks to children into the market. Whatever protections they built in were not enough to prevent these deaths. Litigation will determine whether that was a failure of design, judgment, or corporate priorities.
Your Kid Is Probably Already on These Platforms
Three in four teens have used an AI companion, nearly a third of whom claim it is as satisfying as talking to a real friend. These platforms are engineered to maximize engagement — and in adolescent brains still developing impulse control, that engineering is especially effective. They are built to analyze emotional cues in real time and respond in ways that keep users coming back. Nothing in that design changes when the user is a child in crisis.
The Warning Signs to Watch For
These companies should have built guardrails into their products before putting them in front of children. Until they do, consult with mental health professionals and watch for:
- Sudden withdrawal from family paired with increased phone use
- Distress when a specific app is restricted
- Hiding the screen when you walk in
- References to an online “friend” they’ve never met
- Talking about an AI as though it has feelings
- Mood swings that track with chatbot access
- Any mention of hopelessness or self-harm
What the Companies Knew and When
After the Florida teenager died, Character.AI announced new claimed safety features. Researchers then tested the platform on child-registered accounts and found 669 harmful interactions in 50 hours — one every five minutes. A November 2025 study found that ChatGPT, Gemini, Meta AI, and others consistently fail to recognize mental health crises in young users.
Even now, the industry’s response has been to make chatbots more humanlike, not less. On July 9, 2026, OpenAI launched GPT-Live, a voice model built to speak and listen simultaneously and interject with conversational fillers like “mhm,” explicitly engineered to feel like talking to a real person. OpenAI says it tested the model for risks including self-harm and “emotional dependence on AI,” but the launch came the same month OpenAI was fighting a wrongful-death suit from another family, and just weeks after a California man sued alleging the chatbot worsened a psychotic break. The pattern holds: safety testing announced after the harm is already public, not before the product goes live.
For these companies, engagement is the product and your child’s emotional vulnerability is a revenue stream. They have fought accountability at every turn by lobbying against safety legislation, hiding behind Section 230 and claiming their chatbots are protected speech. That changed in May 2025, when a federal judge ruled in Garcia v. Character Tech. Inc. that a chatbot’s output is a product, not speech.
Change Is Coming, But Not Fast Enough
On April 30, 2026, the U.S. Senate Judiciary Committee unanimously advanced the GUARD Act, which would prohibit AI chatbots from encouraging suicide or self-harm, ban sexual content involving minors, and require age verification for AI companion platforms, with penalties of up to $250,000 per violation. The full Senate has yet to vote on it. New York and California have passed laws requiring chatbots to detect suicidal ideation and connect minors to crisis resources.
What You Can Do
Parents cannot out-engineer a platform built with input from among the world’s best behavioral psychologists. But parents who stay engaged, stay vigilant, and talk openly with their children about the risks of AI chatbots can make a difference. And if your child has already been harmed, you don’t have to navigate what comes next alone.
If your family has been affected by AI chatbot harm, whether through a young person’s mental health crisis, exploitation, or violence connected to AI platform conduct, Aylstock, Witkin, Kreis & Overholtz is here to help. Our attorneys are actively investigating AI harm cases and can help you understand whether you have a legal claim.
Contact AWKO today for a confidential consultation. There is no cost to speak with us, and no obligation.
If you or someone you know is struggling, the Suicide and Crisis Lifeline is available by calling or texting 988.