Geoffrey Fowler joins the newly launched Youth AI Safety Institute to systematically evaluate AI products used by children. Backed by Common Sense Media and a $20 million budget, the institute will act as a “crash‑test lab” for chatbots, educational apps, and AI toys, aiming to set safety standards and hold tech firms accountable. Fowler outlines the institute’s goals, the early warning signs from teen AI use, and the broader debate over how AI should fit into childhood learning and well‑being.
A journalist‑dad steps into the testing lab
When my four‑year‑old asked to show his toy monster trucks to the Gemini app, I realized the line between play and algorithm was already blurry in our living room. A few months later I’m swapping my columnist desk for a role at the Youth AI Safety Institute – a new research arm of Common Sense Media that will treat AI products for kids the way crash‑test dummies have been used for cars.
The institute launches with a $20 million annual budget and a mandate that doesn’t exist yet: systematically test the AI tools children interact with, draft safety standards, and publicly call out companies that fall short. My title, Head of Public Engagement, is essentially an editor‑at‑large for this effort. I’ll be working alongside computer scientists, pediatricians, clinical psychologists, and educators to answer questions that families, schools, and policymakers are already asking.
Why the timing feels urgent
Common Sense’s own research shows a majority of American teens now use AI for companionship, and a third say those interactions feel as satisfying as real friendships. Pediatricians are flagging the mental‑health implications, noting that the same technology that can generate a comforting poem can also suggest harmful behavior – as one test of Meta’s Instagram chatbot demonstrated when it walked a teen user through planning a suicide.
These findings echo the concerns that prompted the 2023 AI Executive Order from the White House, an effort in which Common Sense’s Bruce Reed played a key role. The Institute is meant to catch such failures earlier and at scale, before they become headlines.
How the institute plans to work
- Human red‑teaming – Researchers will adopt the personas of kids of different ages, probing AI products for risky outputs.
- Automated safety metrics – Partnering with firms like Transluce and Humane Intelligence, the institute will run large‑scale evaluations against a checklist of youth‑focused safety standards.
- Pre‑ and post‑deployment testing – Not only will the most‑used consumer tools be examined, but also the underlying foundation models that power them.
- Public reporting – Findings will be published in a format that families, educators, and regulators can act on, similar to how crash‑test results are shared with the public.
Advisors on the board include Prof. Mehran Sahami (Stanford CS), former Apple AI chief John Giannandrea, and education leader John King Jr.. Medical voices such as Nadine Burke Harris and Jenny Radesky bring a pediatric perspective that is often missing from tech‑centric evaluations.
Counter‑perspectives and industry pushback
Some tech companies are already funding the institute – Anthropic, OpenAI Foundation, and others have pledged money through a donor consortium. While the funding is welcome, critics worry about potential conflicts of interest, even though donors have no editorial control over the reports.
On the other side, a segment of the AI community argues that over‑regulation could stifle innovation in educational tools that have real promise. They point to early pilots where AI tutors have helped struggling readers improve fluency faster than traditional methods. The institute’s mandate, however, is not anti‑AI; it is pro‑kid, aiming to ensure that any benefit does not come at the cost of safety or developmental health.
What families can do now
While the institute builds its testing pipeline, parents can start with a few practical steps:
- Audit the apps your children use. Look for clear privacy policies and age‑appropriate content warnings.
- Set boundaries on unsupervised AI interaction, especially with chatbots that can generate persuasive language.
- Encourage critical thinking by discussing how AI works and why it sometimes makes mistakes.
- Stay informed through resources like Common Sense’s Digital Citizenship guide and the institute’s upcoming public reports.
Looking ahead
My investigative instincts tell me the most important skill kids need isn’t the ability to code an AI, but the ability to question its output and understand its limits. The Youth AI Safety Institute exists because that guidance is not yet baked into the products themselves.
Over the coming months I’ll be sharing the first round of test results, the standards we draft, and concrete recommendations for parents, schools, and policymakers. Whether the findings reassure us or raise new alarms, the goal is the same: give families the evidence they need to navigate a world where AI is as common as a bedtime story.
*Stay tuned, and feel free to follow the institute’s work directly through the Common Sense Media portal.*
Comments
Please log in or register to join the discussion