No permanent assurance body certifying AI education products for safety and efficacy

DfE's generative AI product safety expectations, published 2025 and updated 2026, are voluntary self-assessment, with no public register of compliant products and no audit or enforcement behind them. The EdTech Evidence Board pilot at the Chartered College of Teaching, a roughly £800k DfE contract, ended in April 2026 and reviewed supplier-submitted evidence on a voluntary basis. BESA's LendED lists supplier self-declared claims. EEF trials are rigorous and take two to three years per product; Aila's NFER evaluation reports in autumn 2026, two years after launch. Nothing tells a school at market speed whether an AI product used with children is safe and effective.

Schools are buying AI tools now, one procurement at a time, with no trustworthy signal to buy against. Product cycles run in months and the evidence system runs in years. Public money flows to unvalidated products used directly with children.

Make the EdTech Evidence Board permanent and statutory: an assurance body certifying against the DfE safety expectations, with independent audit and a public register, and publishing rapid-cycle efficacy ratings. DfE-funded procurement would be conditional on certification.

// State-led: Instrument: statute making EdTech Assurance Body permanent, with DfE procurement conditionality and audited public register.

Schools buy AI tools for children at market speed, the only evidence pilot lost its DfE funding in April 2026, and the safety expectations remain unenforced and unregistered.