Most chatbot harm cases begin with a download. This one begins with a homework machine. The complaint says the student's only access to any chatbot was the Chromebook her school required her to use, and the school had done the obvious thing: it blocked the chatbot sites. ChatGPT's site was blocked. Chatbot.com was blocked. So was gemini.google.com, Gemini's own front door (paragraph 340).

What stayed open was Search. Google had embedded Gemini inside its search product as AI Mode and AI Overviews. According to the complaint, school administrators could block standalone chatbot sites or disable the Gemini app but could not separately disable Gemini-powered AI Mode and AI Overviews while retaining Google Search. The complaint says she found AI Mode in the autumn of 2025 while doing schoolwork. She had not gone looking for a chatbot, her lawyers say. It appeared inside the tool she was required to use. The complaint also alleges that Google did not alert its K-12 customers that AI Mode and AI Overviews were now an inescapable part of students' school experience (paragraph 26).

The complaint says an initial Search query on February 12 was blocked, but a subsequent query led the student into AI Mode. The filing presents that sequence as evidence that blocking standalone chatbot access did not prevent interactions through Search.

What the logs show, and who could see them

The complaint reports 5,646 logged Chromebook activities from September through November 2025 and 15,871 from December 2025 through February 2026, despite fewer school days in the latter period. Those are activity counts, not verified Gemini conversation or message counts. The filing separately alleges increasingly frequent chatbot use, including late-night sessions.

Some nights, the complaint says, she stayed on the Chromebook as late as 2 a.m. and resumed when she woke (paragraph 339). The filing separately counts the messages themselves: nearly 900 exchanged with Gemini in January and 3,200 in February (paragraph 337). The complaint ties the escalation to December 2025, the month it says Google made Gemini 3 Flash the default model behind AI Mode. At one point, the filing says, she asked the chatbot how to stop her "addiction of asking ai random questions" (paragraph 354).

According to the complaint, school monitoring software flagged a concerning document on March 17, 2026, prompting an urgent notification and parental intervention. The family alleges that Google's products issued no comparable alert. The student survived.

The complaint draws a distinction between Gemini app conversations, which schools can retain or access under applicable settings, and Gemini-powered AI Mode conversations in Search. It alleges that school personnel cannot retain or access student AI Mode transcripts and that Google suppresses those transcripts from student accounts. The filing contrasts this with consumer AI Mode history and Google Takeout export options. These are allegations about access and visibility, not independent proof that Google retains no underlying records.

In the complaint's own words, for student accounts Google "wholly suppresses transcripts of students' chats while in AI Mode" and "hides that those chats ever even happened" (paragraph 238). The family says it still cannot read what the chatbot told her.

A Safety Test Cited in the Complaint

The complaint cites testing attributed to the nonprofit KORA to support its allegation that Gemini 3 Flash handled teen mental-health scenarios poorly. It reports a 7% success figure and a 33% decline against an earlier model. Those figures are the complaint's characterization of KORA's testing, not independently verified measurements in this article: KORA's own published benchmark, read first-hand for this article, scores 32 models from nine providers across 26 child-safety risk areas and places mental-health mishandling among the weakest areas across the industry, but it does not publish the Gemini 3 Flash-specific figures the complaint quotes. Without the original methodology, they cannot establish how Gemini would respond across real-world conversations or whether Google knew of a specific risk to this student.

The complaint goes further, alleging that Google's senior executives knew Gemini 3 Flash was the company's most dangerous model to date and rushed it out in December 2025 anyway. That is the filing's allegation, explicitly untested; the existence of the cited test does not, on its own, establish what any executive knew.

The KORA citation does not stand entirely alone in the coverage. Gizmodo reports that a recent Common Sense Media assessment rated Google's Search AI functions an "unacceptable risk" for children and teenagers, finding that AI Mode and AI Overviews missed clear signs of suicidal ideation. After that assessment, Senators Richard Blumenthal and Josh Hawley wrote to CEO Sundar Pichai about Google's push of AI products to school-age children, Gizmodo reports. Bloomberg's account of the same assessment adds that it found the features normalized disordered eating.

Google's answer

Google has not yet answered in court. In statements to reporters, it has pointed to the safeguards it says already exist. Spokesperson José Castañeda told Reuters the company is reviewing the claims, adding: "We've worked extensively with medical professionals and academic experts to build features that help young people find critical support and trusted resources when they need it." A spokesperson told Gizmodo the company takes the lawsuit "very seriously," and Google says explicit content is filtered for students on Workspace for Education accounts and that it provides age-appropriate AI literacy resources.

The complaint alleges that the student sometimes requested responses without crisis-resource information and that Google's safeguards did not consistently redirect the interactions toward appropriate support. Google says its Search systems provide crisis information for relevant queries and that it monitors attempts to circumvent safeguards. The available record does not independently establish how those protections operated across the student's conversations.

Filed Monday, Google Announced Changes Tuesday

The complaint was filed on Monday, November 2. On Tuesday, November 3, Google announced that Gemini and Google Search would be off by default for K-12 student accounts until school administrators enabled them. Google said schools could enable the services individually, but Search would continue to include integrated AI features. The company attributed the changes to feedback from educators and did not identify the lawsuit as their cause. Google removed the case to federal court on Wednesday, November 4.

The announcement gives schools an additional way to restrict student access to Search, but does not describe an option to retain ordinary Google Search while separately disabling all AI Mode and AI Overview functionality. Google had not specified a complete rollout schedule in the reporting available at publication.

The complaint's eighth cause of action invokes California's Companion Chatbot Law against Google. That raises a distinct legal question: whether the Gemini-powered interactions described in the filing fall within the statute's definitions and requirements, despite occurring through a Search feature rather than a standalone companion-chatbot app. The complaint advances that theory; no court has accepted it.

If you or someone you know is struggling or in crisis, call or text 988 in the United States. Outside the US, visit findahelpline.com to find local support.

Sources

  1. [1] Emma Doe v. Google LLC and Alphabet Inc., complaint (Case 5:26-cv-13205, Document 1-1; state case 26CV507724, electronically filed November 2, 2026; 108 pages; read first-hand for this article, with all cited paragraphs read directly): the filing stamps and eight causes of action (cover); introduction paragraphs 1-4 and 8; the permitted quotation (para. 354); AI Mode and AI Overviews embedded in Search and not separately removable (para. 18); the KORA figures as the complaint renders them (paras. 21-22); no alert to K-12 customers (para. 26); the plaintiff and defendants (paras. 33-35); the Companion Chatbot Law claim's statutory basis (para. 45); the records contrast (paras. 235-239); the message counts (para. 337); the logged activity counts, 5,646 and 15,871 (para. 338); late-night sessions (para. 339); the blocked sites, including gemini.google.com (para. 340); the February 12 search that was blocked and then routed to AI Mode (para. 346).Read source
  2. [2] KORA Benchmark, 'Insights' (read first-hand): an ongoing independent benchmark scoring 32 models from nine providers across 26 child-safety risk areas; mental-health mishandling is among the weakest areas across the industry in its published aggregate results. The public results do not publish Gemini 3 Flash-specific success or decline figures.Read source
  3. [3] EdTech Law Center, 'Emma v. Google' case page and November 4, 2026 press release (the plaintiff's lawyers): the lawyers' account of the claims and of the monitoring alert.Read source
  4. [4] Madrobot, 'Parents sue Google over Gemini on daughter's school Chromebook' (November 5, 2026; read in full): the complaint-based narrative, the eight-claims summary, the crisis-response circumvention passage, and Google spokesperson Jose Castaneda's statement to Reuters.Read source
  5. [5] Gizmodo, 'Lawsuit Claims Student's Google AI Use Turned Into a Life-Threatening Obsession' (November 5, 2026; read in full): Google's fuller statement (crisis hotline information in Search, filtering for Workspace for Education students, jailbreak monitoring, AI literacy resources), the Common Sense Media 'unacceptable risk' rating for Google's Search AI functions, and the Blumenthal-Hawley letter to Pichai.Read source
  6. [6] Insurance Journal (Bloomberg), 'Google AI in School Triggered Teen's Mental Health Crisis, Lawsuit Says' (November 6, 2026): corroboration of the filing and the family's account, and of the Common Sense Media assessment, whose findings Bloomberg reports also included normalized disordered eating.Read source
  7. [7] ALFW, 'Google Lets Schools Turn Off AI. Search Goes With It.' (November 3, 2026): Google's off-by-default announcement for K-12 student accounts and the Search bundling inside it.Read source