DfE Generative AI Product Safety Standards

This alignment statement demonstrates how Skye, Third Space Learning’s spoken AI maths tutor, meets the requirements of the Department for Education’s Generative AI: Product Safety Standards (updated January 2026). It gives schools and Multi-Academy Trusts the assurance they need to procure and use our service with confidence.

13 / 13 standards met 2.3M+sessions 196,000+pupils 4,200+schools 12 yrsexpertise

A closed-loop maths tutor, built for safety by design

Skye is TSL’s spoken AI maths tutor, built on over 12 years of one-to-one tutoring and more than 2,100,000 hours of human tutoring experience. Skye delivers spoken, one-to-one maths sessions to pupils in Years 3 to 6 in primary schools, and KS3 and GCSE in secondary schools, through a secure online classroom. Every lesson is structured around TSL’s National Curriculum-aligned teaching sequences, created and quality-assured by qualified UK teachers. Skye cannot browse the internet, generate free-form content, or deviate from pre-approved curriculum material.

Crucially, Skye is not a general-purpose generative AI tool. It is a closed-loop, pedagogically accurate tutor. This architectural distinction means many of the risks the DfE’s standards are designed to mitigate, such as inappropriate content generation, hallucinations, or open-ended manipulation, are eliminated by design rather than merely managed.

Assessed against all 13 standards. Confident in all 13.

Third Space Learning has assessed Skye against all 13 standards in the DfE’s Generative AI: Product Safety Standards (updated 19 January 2026). We are confident that Skye meets every one. Our compliance rests on three pillars.

Safety by design

Skye’s closed-loop architecture eliminates entire categories of risk at the system level. No open internet access, no free-form generation, no unvetted content.

Robust safeguarding

A multi-layered model: Timeslot Supervisors in every session, an independent tiered red flag monitoring system, full session recording, a dedicated DSO and deputy DSO, and the whole company trained in safeguarding.

Transparency and governance

Comprehensive public policies (safeguarding, data protection, Children’s Code, EU AI Act, GDPR), school-facing DPIAs, annual policy reviews, and a named DSO and deputy DSO. Every claim is backed by published documentation.

Summary alignment with the 13 DfE standards

A concise reference summary of how Skye aligns with each standard. The full statement, with detailed evidence for every standard, is available to download below.

#	Standard	TSL alignment
1	Stated purpose	Clear purpose for Years 3 to 6 and KS3 / GCSE; grounded in EEF (+5 months) and Stanford / NSSA (+0.37SD) evidence; Educate Ventures independent research (92% vs 34% correct); constraints communicated proactively.
2	Educational use cases	Use cases 2, 3 and 4: personalised learning, assessment and analytics, digital AI tutor; diagnostic check-in personalises every session; real-time teacher dashboard.
3	Filtering	Closed-loop architecture prevents harmful content by design (no internet, no free-form generation); independent tiered red flag monitoring system runs in parallel in every session.
4	Monitoring and reporting	Timeslot Supervisors in every session; 5-tier independent red flag alerting system; full session recording and transcription; DSO contact within 24 hours; formal Record of Concern.
5	Security	Cyber Essentials certified; AES-256 plus SSL / TLS; role-based access controls; jailbreak-proof by architectural design; prompt injection-resistant.
6	Privacy and data protection	UK GDPR compliant; no data used for AI training; student-facing privacy policy; DPIA support for schools; ICO Children’s Code compliant; DPO named.
7	Intellectual property	No pupil or teacher IP used for model training or improvement; defined retention schedule (end of third academic year after last session).
8	Design and testing	Safety by design; co-designed with schools; internal testing suite for simulated interactions; feature flags for controlled rollout; no emotion recognition or subliminal techniques.
9	Governance	Named DSO (Bryan Tookey, Level 3), Deputy DSO (Jenni Hoy) and Board lead (Tom Hooper, CEO); quarterly safeguarding updates to Board; published trust and compliance page; formal complaints process.
10	Cognitive development	Structured I Do / We Do / You Do pedagogy prevents cognitive offloading; scaffolded hints not answers; diagnostic check-in and check-out plus confidence check-out verify genuine learning.
11	Emotional and social development	Robot avatar; task-oriented language; redirects personal disclosures to a trained adult; no relationship cultivation; timetable-based session limits; home usage visible in reporting dashboard.
12	Mental health	Verbal distress detection with structured escalation pathway; immediate alert to Schools Team plus direction to Timeslot Supervisor; DSO review within 24 hours; formal safeguarding log; annual staff training.
13	Manipulation	No subliminal techniques, emotion recognition, dark patterns, or ads; ad-free platform; low-stakes transparent rewards only; sessions do not run long artificially.
All 13 DfE Generative AI Product Safety Standards