Britwise is an education provider using AI for language assessment + tutoring. Under the EU AI Act (Reg. 2024/1689), this qualifies as a high-risk AI system (Annex III, point 3 — education and vocational training). We comply with the corresponding obligations.

OpenAI GPT-5.2 / GPT-4o (assessment, lesson generation) · Anthropic Claude Haiku 4.5 / Sonnet 4.5 (Angie chat tutor, content moderation) · Deepgram Nova-3 (speech-to-text, EU endpoint) · ElevenLabs Flash v2.5 (text-to-speech, 8 British accents) · Hume Octave (voice prosody for confidence scoring).

• IELTS band predictions are not used as a substitute for the IELTS exam. We surface predicted bands but explicitly mark them as estimates. • Safeguarding (KCSiE) flags route to a human Designated Safeguarding Lead within 30 min / 2 h SLA. AI never makes safeguarding decisions autonomously. • HR ROI dashboards are decision-support tools — promotion / role decisions remain human.

We test pronunciation grading for fairness across non-native accents quarterly. Current findings: minor accent bias (≤0.2 CEFR band) detected for South Asian and West African accents, mitigated by ensemble grading. Full audit report available on request to dpo@britwise.school.

• Right to explanation: any AI score includes a written rationale. • Right to human review: contest any AI score by emailing support@britwise.school. • Right to opt out: minors can opt out of voice recording in B2S deployments (parental consent gate).

All AI-generated content is labelled as such in the UI (Angie chat, lesson summaries, parent reports). We do not pass off AI output as human-authored.