The Next News

// technology and science news, explained

AINews3 mins read3 sources

England proposes safety certification for AI tutors, chatbots and other edtech used in schools

The Department for Education proposes independent checks on student-facing AI, web filters and high-risk data processing. Its consultation runs to February 4.

By AI Desk ·

England’s Department for Education (DfE) has proposed a scheme under which independent assessors would certify the safety of education technology, including AI tutoring tools and chatbots built for children. A consultation opened on November 10 and closes at 11:59 p.m. on February 4, 2027. The department plans to publish its response in spring 2027.

The scheme would cover safety, not teaching quality: certification “would not make a wider assessment of a product’s overall quality or educational effectiveness,” the consultation document says. In its foreword, Paul Waugh, the Minister for Early Education, writes that the aim “is not to slow innovation.”

Four kinds of product in scope

DfE names three main areas of risk for children using edtech: data infringements, exposure to harmful content, and effects on their cognitive, social and emotional development and mental health. It proposes four product categories, covering products that settings use or make available for children aged 0 to 18.

The four product categories in DfE’s proposed edtech certification scheme, and what the consultation says each would be assessed on
CategoryWhat the consultation proposes
Filtering productsBlock harmful content, apply the Internet Watch Foundation and Counter-Terrorism Internet Referral Unit blocklists, and block VPNs and proxies. DfE asks whether filters should also handle AI-generated content.
High-risk processing of children’s dataLikely to include AI tutoring tools, personalized learning systems, behavior and safeguarding systems, platforms for special educational needs and disabilities, wellbeing and mental health platforms, and school management information systems. Suppliers would show a data protection impact assessment and compliance with ICO codes.
Student-facing generative AI toolsAssessed against eight of DfE’s generative AI product safety standards, from filtering and monitoring to cognitive development, mental health and manipulation.
Student-facing personalized learning toolsA modified version of the generative AI standards, aimed at design features that add pressure, gamify learning or prolong engagement “without pedagogical justification.”

Left out would be static content such as PDF textbooks, hardware, and office and staff-facing tools, including generative AI tools for staff, that do not carry out high-risk processing of children’s data. Technical monitoring products would not be covered at launch, because DfE says their safety depends largely on how staff act on their alerts.

The product, not the model

For generative AI, certification would attach to the app, chatbot or tutoring tool that pupils use, not to the underlying model or API. The supplier would be responsible even when its product runs on a third-party model, because it controls how the model is integrated and configured, the document says. AI tools or “agents” that a teacher or school builds for use only in that setting would be exempt.

The AI tests already exist as guidance. DfE published generative AI product safety standards for edtech suppliers on January 22, 2025, and added standards on cognitive development, emotional and social development, mental health and manipulation on January 19, 2026. Assessment could combine documentation and technical reviews, benchmarking, safeguarding scenario tests and red teaming, the consultation says.

AI tools would also be checked more often. DfE suggests full recertification every three years for the other three categories, with annual surveillance checks in between, but says the pace of change means generative AI tools may need recertifying more frequently, “for example, every year.”

Who would run it, and whether schools must use it

The United Kingdom Accreditation Service (UKAS) would accredit conformity assessment bodies against the international standard ISO/IEC 17065, and suppliers would pay those bodies to assess their products. DfE would keep a public register of certified products. Of three ownership models, it prefers to own the scheme and license its delivery to one or more outside organizations.

Certification would not be compulsory at first. DfE says it “does not have specific powers” to require it, so it would set an expectation in statutory guidance, including Keeping Children Safe in Education, while letting settings rely on other credible evidence. It asks whether legislation may be needed later, and estimates that setting up the scheme would take at least 18 months before any product could be certified.

The consultation also asks suppliers what annual certification cost they could absorb without passing it on, with options from under £2,000 to more than £10,000. DfE says schools want the assurance: in its Technology in Schools Survey 2025, half of school leaders said safeguarding and data concerns were to some or a great extent a barrier to increasing schools’ uptake of technology.

Sources

  1. Introducing an EdTech and AI product safety certification scheme Department for Education (GOV.UK) gov.uk
  2. Introducing a certification scheme for assuring EdTech and AI product safety (consultation document) Department for Education consult.education.gov.uk
  3. Generative AI: product safety standards Department for Education (GOV.UK) gov.uk

Checked against the 3 sources above on November 15, 2026. Spotted an error? Tell us.