Google wants to expand the use of artificial intelligence in education. Two United States senators have written to its chief executive asking what the company knows about the effects of that on children, and have set a deadline of 2 November for the answers.
Senators Richard Blumenthal and Josh Hawley sent Sundar Pichai a letter raising concerns about children’s mental health and educational development, and asking for explanations on nine specific points.
The nine
1. What research has Google conducted into the possible negative effects of its AI products — Gemini, AI Overviews, AI Mode and others — on children’s mental health, and what did that research find?
2. Why does Google leave AI features switched on by default on devices used in schools, and why, in some cases, is there no option to turn them off?
3. Have school authorities, teachers or parents raised concerns with Google about students’ dependence on AI for study or about effects on mental health — and if so, what did the company do about them?
4. What support has Google given to institutions that reported students being harmed by its AI products?
5. What promotional programmes has Google run to expand AI use in education, and how much has it spent on them?
6. What guidance and supporting material has it given to institutions, administrators and teachers — specifically on appropriate use of AI and on protecting students’ wellbeing?
7. Has any child been encouraged into harmful behaviour through Google’s classroom AI — and has the AI failed to detect signs of suicidal intent or self-harm?
8. Following the findings of Google’s own researchers in the paper The Ethics of Advanced AI Assistants, what steps has the company taken to reduce the potential harms of AI to children?
Why the seventh question is the one that matters
Most of the letter asks for disclosure. Question seven asks for something harder: an admission of a specific failure, in a specific category, that the company would have had to notice internally in order to answer honestly.
It is also the question that explains the rest. A chatbot in a classroom is not only an answering machine; it is a thing children talk to, sometimes about what is going on with them. Whether it recognises a child in distress, and what it does about it, is a safety question rather than an educational one — and it is being asked of a product that, by question two’s account, is switched on by default and in some cases cannot be switched off.
What this means beyond the US
None of this is binding anywhere else. But the answers, if Google gives them, will be the most detailed public account anyone has of what the company has measured about AI and children — and the same products are in use in schools far outside the senators’ jurisdiction.
For parents and schools here, the practical version of question two is worth borrowing: on the devices your children use, which AI features are on by default, and can you turn them off?
What a deadline of this kind is worth
A senator’s letter is not a subpoena. Google can answer narrowly, or late, or decline. What the deadline does is create a dated record of the question having been asked — which matters later, if a specific case ever reaches a court or a committee with the power to compel.
That is the usual function of letters like this one, and it is why the wording of question seven is careful. It asks about failures the company would already know about, not about failures it would have to go and look for.
