Independent assessments and a public register would guide purchasing decisions, with proposed checks covering children’s data, AI interactions and potentially harmful product design
The DfE is consulting on safety certification for higher-risk edtech and AI products used in education settings in England. Conceptual illustration.
The Department for Education (DfE) is consulting on a safety certification scheme for edtech and AI products used in England’s schools, colleges and early years settings, proposing independent assessments to help education providers decide which technologies to buy and use.
The scheme would cover four higher-risk product groups: filtering systems, technology conducting high-risk processing of children’s personal data, student-facing generative AI tools and student-facing personalized learning products. Certified products could display a safety badge and appear on a publicly available register maintained by the DfE.
Certification would become the normal route for demonstrating that relevant products meet safety standards under proposed changes to statutory safeguarding guidance. However, the scheme would initially be non-mandatory, with education settings able to rely on credible equivalent evidence where justified.
The consultation opened on November 10, 2026, and closes on February 4, 2027. It seeks views from education providers, suppliers, assessment organizations, safeguarding specialists, parents and students on the scheme’s scope, assessment criteria, costs and implementation.
Safety certification would shape school purchasing
The DfE proposes introducing expectations through Keeping Children Safe in Education, England’s statutory safeguarding guidance for schools and colleges, with corresponding expectations in the Early Years Foundation Stage statutory framework.
When buying or using higher-risk products within the scheme’s scope, settings would be expected to obtain proportionate assurance that the technology meets relevant DfE safety standards. Certification would ordinarily provide that assurance.
An alternative route would remain available. Settings could use other credible evidence, explaining why it provides equivalent assurance and identifying any additional safeguards needed for the intended use.
The consultation gives technical, vocational and applied learning as examples where using a non-certified product could be justified. Older students may need experience of technologies used in workplaces, provided the setting can demonstrate appropriate safety protections.
Ofsted and other inspection bodies could have a role through their consideration of safeguarding arrangements. The DfE says this would focus on leaders’ decisions, safeguarding culture and effects on children, rather than holding settings responsible for product failures beyond their control.
Certification could also be incorporated into procurement frameworks, initially by identifying certified products and potentially becoming a condition for inclusion in relevant frameworks.
The department states that it does not currently have specific powers to require settings to use certified edtech or suppliers to obtain certification. The consultation therefore asks whether future legislation would be needed to make certification mandatory in defined circumstances.
The badge would assess safety, with educational effectiveness largely outside scope
The proposed scheme would assess whether products meet defined safety requirements, including protections against emotional and developmental harm, appropriate handling of data, system security and accurate representation of safety-related functions.
It would not provide a general endorsement of teaching quality or learning outcomes.
"Certification would not make a wider assessment of a product’s overall quality or educational effectiveness," the consultation says.
The exception would be where effectiveness directly affects safety, such as a product failing to perform a function intended to protect children.
The DfE identifies three main areas of potential risk: unlawful or inappropriate use of children’s data, exposure to harmful content, and effects on cognitive, social and emotional development and mental health.
It also cites the Technology in Schools Survey 2025, in which 50% of leaders said safeguarding and data concerns were, to a great or some extent, a barrier to increasing technology use in schools.
The department intends to continue supporting decisions about educational effectiveness through separate evidence and guidance, including its Edtech Testbeds program with the Education Endowment Foundation.
Four product categories, with exemptions for lower-risk tools
The proposed scope extends to relevant products made available to children in early years, schools and further education in England. A product could fall into more than one certification category.
Potentially covered products include AI tutors, safeguarding systems, platforms supporting special educational needs and disabilities (SEND), and school management information systems.
Lower-risk exclusions would include static digital textbooks, hardware, and staff-facing tools that do not conduct high-risk processing of children’s personal data.
Teacher- or school-created AI agents and custom tools used only within that setting would also be exempt. The setting would retain responsibility for their safe and lawful use.
Technical monitoring would have no standalone certification category initially, although monitoring products could fall within the high-risk data processing category.
Generative AI checks would cover emotional dependence and cognitive development
Student-facing generative AI products would be assessed against eight areas drawn from the DfE’s existing Generative AI: product safety standards.
These cover filtering, manipulation, mental health, emotional and social development, cognitive development, governance, design and testing, and monitoring and reporting.
Products would need to demonstrate measures to prevent harmful content and manipulative or exploitative design. Assessments would also examine protections against emotional dependence, potential harm during mental health emergencies, and cognitive deskilling or longer-term developmental harm.
Governance requirements would include risk assessments, formal complaints mechanisms and understandable information about operations, decisions and data handling. Design and testing requirements would cover child-centered development, mitigation of identified risks and testing with relevant stakeholders, including children.
Monitoring requirements would include logging prompts and responses, analyzing performance and alerting local supervisors when harmful or inappropriate content is accessed or attempted.
Certification would identify the product version, intended age range, relevant configuration, assessment methods, known limitations and conditions of use.
It would apply to the student-facing application, rather than an underlying model or API alone. Suppliers integrating third-party models would remain responsible for their products.
All student-facing generative AI products would also be assessed under the high-risk data processing category. The DfE proposes avoiding duplicate assessment of data requirements across the two strands.
The department is considering further standards on privacy-preserving reporting of product use and safety concerns, alongside clearer expectations where students’ work is used to develop or improve AI tools.
Personalized learning tools face scrutiny over gamification and engagement
The proposals extend beyond products that generate AI content. Student-facing personalized learning tools without generative AI would be assessed against adapted standards covering design and testing, governance, cognitive development and manipulation.
The DfE wants to examine features that create unnecessary pressure or prolong engagement without an educational justification. The consultation specifically invites evidence on rewards, streaks, leaderboards, notifications, reminders, recommendation systems, social interaction and approaches to feedback.
It also asks how these features affect children with SEND, including the role of accessibility and additional support.
The department acknowledges that the evidence base is still developing. It is seeking a proportionate approach that addresses harmful design while allowing engaging educational products that teachers and students value.
Detailed standards for this category would be developed with stakeholders before implementation.
Filtering requirements could extend to AI-generated content
Filtering products would be assessed against criteria drawn from existing school and college filtering and monitoring standards.
Suppliers would need to demonstrate that their products block harmful content without unreasonably obstructing teaching, learning or administration. Requirements would include separate student, staff and guest profiles, protection across managed devices taken off-site, and network-level filtering for guest and personally owned devices using school networks.
Other proposed checks cover protected administrative logs, relevant security updates, multilingual content detection, customizable alerts and resistance to methods used to bypass filtering.
Products would need to integrate relevant Internet Watch Foundation and Counter-Terrorism Internet Referral Unit blocklists, with safeguards preventing users, including system administrators, from disabling or altering those protections.
The consultation also proposes restrictions on using data collected for filtering for other purposes, including profiling children or family members who share a device.
A separate question asks whether certification should require effective filtering of real-time, dynamic, personalized and AI-generated content. If adopted, that requirement would apply when the scheme launches.
High-risk data processing would require evidence of legal compliance
For products conducting high-risk processing of children’s personal data, certification would focus on demonstrating compliance with applicable data protection requirements.
Suppliers would need to provide evidence of a Data Protection Impact Assessment and action taken to address identified risks. Proposed criteria also cover relevant intellectual property requirements, compatibility with the DfE’s cybersecurity standards and suitable organizational security measures, such as Cyber Essentials Plus certification.
The DfE intends the planned ICO EdTech Code of Practice to inform certification requirements once finalized and implemented. Other relevant codes would apply according to the legal and regulatory framework in force when the scheme is established.
The Information Commissioner’s Office would retain its role as the independent data protection regulator. Certification would provide an additional layer of assurance, without replacing regulatory enforcement or providing immunity from action.
The department also says it wants to avoid duplicating evidence already required through its management information systems framework.
Independent assessors would certify products
The DfE’s preferred model is to own the scheme and license delivery to third-party organizations, retaining oversight of standards and governance.
Independent conformity assessment bodies, known as CABs, would assess products and make certification decisions. The United Kingdom Accreditation Service (UKAS) would assess whether those bodies have the competence and impartiality to carry out the work.
Suppliers would apply to an accredited, appointed assessment body and pay a certification fee. Where several bodies cover the same product category, suppliers could choose between them.
Assessment could combine documentation reviews, technical examination, benchmarking, realistic safeguarding scenarios and red teaming, in which testers deliberately challenge or attempt to misuse a product to identify failures.
Suppliers relying partly on platform or institution-level safeguards would need to show how those controls work and how the relevant safety requirements are consistently met.
For products still being developed or tested, the DfE is considering a staged readiness route based on self-assessment or documentation review. That would not itself confer certification.
The consultation asks when full assessment should become the normal requirement: before first use in a school or college, after a specified number of settings or users, or according to risk.
Successful products could display a badge, with the public register showing information such as the assessed product categories, certifying body and certificate expiry date.
Recertification and withdrawal would provide ongoing checks
For filtering products, high-risk data processing and personalized learning tools, the DfE suggests full recertification every three years, supported by annual targeted checks.
Student-facing generative AI tools could require more frequent recertification, for example annually, because of the pace of technological change and the risks involved.
Suppliers would need to notify assessors of material product changes. The assessment body would determine whether a change requires focused reassessment or full recertification.
Assessment bodies would also investigate complaints and reports of non-compliance. Certification could be suspended or withdrawn where a product fails requirements or a supplier does not complete corrective action.
Settings would be notified. Depending on the circumstances, they could be advised to stop using a product immediately or continue temporarily with safeguards where abrupt withdrawal would create a greater safeguarding or operational risk.
A product losing certification would be removed from the register, and its supplier would lose the right to use the scheme’s badge.
Supplier costs and support remain open questions
Certification fees have not been set. They would be determined by assessment bodies and depend on the final criteria and assessment process.
The consultation asks suppliers how much annual certification cost they could absorb before passing at least some of it on to customers. It also seeks evidence on indirect costs, including gathering documentation, changing products, preparing for assessment and maintaining compliance.
Possible support for smaller suppliers and new entrants includes templates, self-assessment tools, technical assistance, reduced fees or financial support, and additional preparation time. These are options being tested through the consultation.
Suppliers are also asked how much lead time they would need for each product category and which existing standards or assurance processes they already meet.
The DfE says implementation would take account of procurement and contract cycles, with a transition period for suppliers and education settings.
Responses due by February 4
The consultation is open to everyone, including school and college staff, early years providers, suppliers, safeguarding and SEND organizations, researchers, parents, carers and students. Most questions are optional.
The DfE intends to publish its response in spring 2027. Establishing the scheme would take at least 18 months before products could be certified, followed by a transition period.
Responses must be submitted by February 4, 2027.
Read the consultation and submit a response
ETIH Innovation Awards 2027 open for entries
The ETIH Innovation Awards 2027 are now open for entries, with 26 categories recognising companies, institutions, projects and individuals making a measurable impact across education technology. The global awards span AI in education, digital learning, higher education, skills and workforce, inclusion, education operations, evidence and impact, with entries managed through ETIH’s new online awards platform.
Entries are open worldwide, with the early-bird deadline on Friday 18 December 2026 and the final entry deadline on Friday 26 February 2027 at 23:59 UK time.
Explore the 2027 categories and judging criteria, read the How to Enter guide or start an entry through the ETIH Innovation Awards 2027 platform.