A 13-Year-Old Asked AI About Suicide on Her School Laptop. It Didn’t Get Flagged for Weeks

According to a lawsuit filed this month, a 13-year-old spent about three months on her school-issued Chromebook talking with Google's AI about depression and suicide, then drafted her own eulogy, before monitoring software prompted the school to alert her parents. Her mother then watched her write a suicide note in real time and stepped in. The family's lawyer says the alert first went to the personal phone of a counselor who had left the school nine months earlier.

The lawsuit alleges Google was negligent and that the chatbot "coached" the girl. Those are allegations, and Google says it is reviewing them and that its systems aim to prevent harmful conversations. But the timeline exposes something that doesn't depend on how the case turns out: schools are layering AI into search, then trusting monitoring tools that may not see it.

We want to handle this carefully, because a child nearly died and the facts are still contested. What the suit alleges is that the school blocked the standalone Gemini site, but the girl reached the chatbot anyway through AI Mode inside Google Search. WIRED reports she continued interacting with it thousands of times over weeks. If that's accurate, the school's block was a fence with a gate built into it.

Our angle is the stack of safety systems that each assumed another one was working. The AI is supposed to steer at-risk users to help. The school's filters are supposed to block risky tools. Third-party monitoring, possibly from GoGuardian or Gaggle according to the family's lawyer, is supposed to flag danger. And the alert is supposed to reach someone who can act. In this account, the first two leaked, the third took almost two months after explicit questions, and the fourth went to an ex-employee. One conscientious former counselor passing a message along is what stood between this family and a funeral.

We advocate for AI in education when it helps students learn. That case gets much weaker when vendors make AI hard to turn off. Just last week, parents and teachers objected to Google requiring schools to turn on generative AI features in order to use Search, and Senator Ed Markey cited this lawsuit in a letter demanding schools be allowed to disable AI Mode and AI Overviews. When administrators can't fully control a feature, they also can't fully monitor it. A school-issued device has to be a controlled environment, or it isn't a school device in any meaningful sense.

We're also not going to pretend surveillance is the answer. The main evidence that these monitoring tools work comes from educators who believe they help, in a survey sponsored by Gaggle, and civil rights groups have warned they erode the trust between students and the adults who might help them. Rand's Benjamin Boudreaux puts it well: schools may have rushed into tools without building the human process to act on alerts. A keystroke scanner doesn't replace a counselor who knows a child.

If this pattern scales, conversational AI becomes the place kids take their darkest thoughts, inside products their schools can't switch off and their monitors can't read. That's where this leads unless three things change. AI products for minors need hard escalation paths in a crisis. Schools need the ability to disable AI features. And every alert system needs a tested, current human on the other end.

The case, first reported by The New York Times, is one of several lawsuits alleging harm from AI chatbot use, including claims involving delusions and suicide. WIRED identified the school, which did not respond; it has since added suicide-prevention links to its website. In the US, anyone in crisis can call or text 988 to reach the Suicide & Crisis Lifeline.