TL;DR
On 10 November the DfE (Department for Education) opened a consultation on certifying the safety of AI and EdTech products that English schools, colleges and nurseries use. Independent assessors would certify products against minimum safety criteria. Responses are due by 4 February 2027, and DfE says the scheme would take at least 18 months to set up.
What would be covered
DfE proposes a risk-based scope of four product categories:
- filtering products
- EdTech whose handling of children’s data is high-risk
- generative AI tools that students use directly
- personalised learning tools that students use directly
The generative AI category takes in AI tutors, chatbots and content-generation tools built for children. The document argues that because their output is produced in real time and is unpredictable, existing filtering and monitoring may not be enough to protect children from harmful content. It also flags harm to children’s development and mental health from badly designed tools.
Lower-risk products such as email, word processors and static PDF textbooks fall outside the proposal. So, at launch, do tools whose main job is safeguarding monitoring, because their value depends on how staff act on the alerts.
How it would work
Assessment would fall to conformity assessment bodies holding UKAS accreditation, and suppliers would pay them a fee. DfE says it expects fees to be “reasonable and proportionate” and asks suppliers how much cost they could absorb.
The scheme would not be compulsory. DfE would write an expectation into statutory guidance such as Keeping Children Safe in Education, yet it has no power to make settings buy certified products. The consultation asks whether legislation should follow, noting that any mandate would need a new legal basis. DfE expects to publish its response in spring 2027.
What it means
If it goes ahead, English schools would get a national safety check on the AI products they buy. DfE has itself tendered for AI tutoring for disadvantaged pupils, and on 9 November more than 300 signatories urged a pause on AI tutors. Separately, BSI launched three AI testing and certification schemes for UK firms in October.
In our view, the voluntary design is the weak point. With an 18-month build and a transition period after that, AI tools will be in classrooms for years before any certificate exists, and schools will still need their own checks in the meantime. Suppliers should respond on fees: a cost that small vendors cannot absorb could narrow the market rather than make it safer.