Ahead of the 2025-26 school year, the education-technology company GoGuardian quietly expanded what its safety software can see. Beacon, an AI monitoring tool already installed on millions of school-issued Chromebooks and Windows laptops, began scanning not just students’ emails, documents, and browsing history but what they type into ChatGPT, Google’s Gemini, and the chatbot app Talkie-AI — flagging language that suggests a student may be at risk of harming themselves or someone else. A year into that expansion, the feature sits at the center of a familiar argument in education technology: is this protection, or surveillance with a friendlier name?
What Beacon Now Sees
GoGuardian announced the AI-chat visibility feature on September 17, 2025, timed to National Suicide Prevention Month, after rolling it out over the summer. The company says Beacon uses natural language processing to detect what it calls nuanced signals of self-harm or harm to others inside a student’s generative-AI conversations, then sends an alert to whoever a school has designated to respond. GoGuardian says Beacon is now used by more than 51,000 educators, part of a installed base the company has previously said spans more than 10,000 schools and 25 million students nationwide. “Student safety can’t wait,” said Brian Larkin, GoGuardian’s Director of Product Management, in the announcement. “Schools need tools that evolve as fast as student behavior does.”
The Case for Watching
GoGuardian frames the expansion as a response to a staffing crisis, not just a technology opportunity. Citing National Center for Education Statistics data, the company says more than half of public schools report insufficient mental health staffing or funding, while 58% say they’ve seen a rise in students seeking mental health support. “Every child deserves the chance to be safe, seen, and supported,” said CEO Rich Preece. The product itself has a grim origin story: Axios has reported that Beacon’s self-harm detection emerged after a student died by suicide at a school piloting the software, prompting an administrator to ask whether better monitoring could have caught it.
The Numbers GoGuardian Points To
The company’s strongest evidence comes with caveats attached. An April 2025 medRxiv preprint — not yet peer-reviewed — found youth suicide rates were 26% lower between 2021 and 2022 in counties actively using Beacon, after adjusting for demographic and regional differences; that’s a correlation, not a controlled trial. GoGuardian also says Beacon alerts notified schools about 18,623 students at risk of harm between March 2020 and July 2024, language the company stretches into a claim that it “prevented” that many students from physical harm. Its most vivid example is Neosho School District in Missouri, which it says once recorded youth suicide rates 30 times the national average and has reported zero suicide deaths in the four years since adopting Beacon alongside separate community mental-health initiatives.
Why Critics Call It Surveillance, Not Safety
Axios reporting on the software has surfaced a harder problem: accuracy. The outlet found Beacon can stumble in both directions — flooding schools with unnecessary alerts that bury real warnings, or missing genuine signs of danger entirely. GoGuardian’s own CEO has acknowledged the system is tuned to err toward over-alerting, but the company has declined to publish failure-rate statistics for its software. That opacity matters at scale: a March 2026 Education Week report, citing data from competitor Securly, found that roughly one in five student interactions with generative AI on school devices involved cheating, self-harm, bullying, or other problematic behavior — a volume of flagged activity that risks overwhelming the humans meant to review it.
A Wider Shift Across School Tech
The urgency behind these tools is partly a function of how fast students have adopted AI that schools never deployed themselves. A Jed Foundation report released in July 2026, drawn from survey data collected between October 2025 and April 2026 across more than 5,500 middle and high schoolers in 21 U.S. schools, found that four in five students already use some form of AI — usage that schools can only partially see, since it happens on personal devices and consumer apps GoGuardian’s extension doesn’t reach. That gap is pushing rivals like Securly toward similar NLP-based detection, turning keyword-matching into an industry-wide bet on machine learning.
What’s Next
GoGuardian is currently offering districts a free 90-day Beacon trial, a sign it expects adoption to keep expanding into the 2026-27 school year. The Jed Foundation has said its full findings on teen AI use will land this fall, which should sharpen the policy debate considerably. The central question reporters and researchers keep returning to isn’t whether these tools can technically flag concerning language — it’s whether schools, parents, and regulators are equipped to judge whether the alerts are accurate, proportionate, and worth the privacy trade-off, especially as the technology being monitored evolves faster than any school district’s policy manual.
Photo: Adam Sondel / PEXELS via Pexels
