A rating body aimed at AI products, not just screen time
Common Sense Media built its reputation rating movies, TV, and apps for age appropriateness, a familiar but limited lane. The Youth AI Safety Institute is a different kind of organization: it will develop child safety standards with clinicians and child development specialists covering learning, attention, social-emotional development, and wellbeing, then test specific AI products against those standards and publish the results. The organization is explicit that results will be public, available to families, educators, policymakers, and industry, not kept inside the companies being assessed.
Robbie Torney, the institute's Head of AI and Digital Assessments, compared the goal to established consumer protections for children such as car seats and toys: independent testing against shared standards, not self-certification. That comparison is the institute's whole thesis. Car seats are not safe because manufacturers say so. They are safe because an independent body tests them against a public standard and publishes the result. Common Sense Media is betting that AI products aimed at children need the same infrastructure, and that nobody else was building it fast enough.
Why Murthy is the chair, not just a name on a press release
Vivek Murthy served as US Surgeon General twice, from 2014 to 2017 and again from 2021 to 2025, and his second term produced a widely cited advisory connecting social media use to youth mental health. He joined Common Sense Media's board in October 2025 and now chairs the new institute. His framing of the problem is specific and worth quoting directly: the risk AI poses to kids and families is a today problem, not a tomorrow problem, and he has argued that fixation on extreme future AI risk should not crowd out attention to harms already happening.
That framing puts Murthy in direct tension with a large share of AI safety discourse, which spends most of its energy on frontier model risk, existential scenarios, and model capability evaluations. Murthy is instead pointing at the products already in millions of children's hands right now: AI companions, homework helpers, and search tools. For school district technology leaders fielding parent complaints about chatbot use today, that is a far more immediately useful lens than most AI governance conversations currently offer.
The usage numbers that justify the urgency
The institute is leaning on its own 2026 Generation AI research to make the case that this cannot wait. Fifty-nine percent of children and teens use AI to search for information or facts, and 55 percent use it for homework help, both now mainstream behaviors rather than edge cases. Separate 2025 research from the same organization found more than half of American teens regularly chat with an AI companion, a category of product that barely existed in a mainstream form three years ago and now sits alongside homework tools as a default use case.
A further statistic, drawn from the Center for Democracy and Technology and cited by the institute, found 42 percent of students use AI for mental health support, as a friend, or to escape from real life. That figure is the one that should concern school technology leaders most directly, because it describes emotional dependency, not productivity use, and it sits entirely outside the scope of most district AI acceptable-use policies written around plagiarism and academic integrity.
Funded by the companies it will grade
The institute's funding list includes Anthropic, the OpenAI Foundation, and Pinterest, alongside philanthropic funders including the Bezos Family Foundation, the William and Flora Hewlett Foundation, and the Walton Family Foundation. Common Sense Media is explicit that some of its funders make AI products the institute may test, and it states that it retains full responsibility for standards, research, and evaluations along with editorial independence over published findings. That disclosure is honest, but it does not eliminate the structural tension of a rating body funded in part by the industry it rates.
The advisory group is built to lend credibility against exactly that criticism. It includes John Giannandrea, a former Apple AI executive and former Google search and AI chief, John King Jr., the SUNY Chancellor and former US Education Secretary, Nadine Burke Harris, California's former Surgeon General, and researchers from the University of Michigan Medical School and Stanford's computer science department. Partner organizations include Stanford Medicine Brainstorm, Harvard's Berkman Klein Center, and Crisis Text Line. That is a roster built to survive scrutiny from both AI industry critics and child safety skeptics simultaneously.
What this means for vendors and district buyers
Any company selling AI products into K-12 or consumer markets where minors are a meaningful user base should expect a Common Sense Media scorecard to become a reference point the way its media ratings already are for parents and some school procurement checklists. A public, independently produced risk assessment is a different kind of pressure than a regulatory investigation or a lawsuit, because it is designed for immediate, repeated consumption by exactly the parents and school boards who decide whether a product gets adopted or banned district by district.
District technology leaders evaluating AI tools for classroom or student use should watch for the institute's first published standards and treat them as one input alongside existing privacy and accessibility reviews, feeding into the district's own vetting process rather than substituting for it. The institute's standards carry no legal force, but reputational force in K-12 procurement often moves faster than regulation does, and a bad public score from a trusted, clinician-backed rating body is the kind of risk that gets a contract pulled from a school board agenda before any formal compliance review is even scheduled.
Preparing for the first scorecards
Vendors should start preparing for this scrutiny now rather than after the first scorecards land. That means documenting age-appropriate design choices, companion-style engagement limits, and crisis-response behavior in the same detail already expected for FERPA and COPPA compliance, and being ready to walk a reviewer through those choices on request. The advisory board includes a former Apple AI chief and a sitting university chancellor, reviewers who will recognize a prepared answer from a scripted one within a few follow-up questions.
The institute plans to run this testing on an ongoing basis rather than as a one-time report, which means a product's standing can change with each model update or feature release. Vendors that build a repeatable internal review process now, tied to the same categories the institute has already published, research, standards, product evaluation, and publication, put themselves in a position to respond to a request for evidence in days rather than scrambling to assemble documentation after a scorecard has already gone public.
