Responsible AI
Last updated: 11 May 2026
How MimWise classifies under the EU AI Act
MimWise uses AI to evaluate learning outcomes and to steer the learning process. This places it in the high-risk category for educational AI under Annex III, point 3 of the EU AI Act (Regulation (EU) 2024/1689). As a provider/deployer of a high-risk system we are committed to meeting Articles 9 to 17 (risk management, data governance, technical documentation, record-keeping, transparency, human oversight, accuracy, robustness and cybersecurity) and Article 27 (Fundamental Rights Impact Assessment) where applicable.
Transparency (Article 50)
- Interacting with AI: every page that calls an AI model carries an in-product banner stating it is AI-generated.
- Synthetic content labelling: AI-generated audio (Piper, gTTS, edge-tts) and AI-generated diagrams will be marked with a machine-readable provenance signal (C2PA / watermark) by 2 August 2026 as required by Article 50(2).
- Model identification: each AI feedback record stores the engine name (e.g.
groq_http,mistral_http) and the underlying model identifier so you can request the lineage of any given output.
Human oversight
Every AI-generated assessment carries a “Request human review” action. Selecting it sends a request to a MimWise reviewer who responds within 30 days. You can also contest a decision under UK GDPR Article 22 / UK DUAA 2024 s. 80 by emailing info.mimwise@gmail.com.
Accuracy and limitations
AI feedback is a study aid, not a replacement for a teacher. Models can be wrong, may hallucinate references, and may reflect biases present in their training data. Always cross-check important facts with a primary source. AI-generated grades should never be the sole basis for a high-stakes academic decision; if an institution wishes to use MimWise that way, it must put its own Article 22 safeguards in place and accept joint-controller responsibilities.
Bias and fairness
- We monitor model outputs for systematic differences across language, learning style and disability flags.
- We provide multiple persona rubrics (“Prof Leniency”, “Prof Precision”) so learners are not locked into a single feedback style.
- We invite users to report concerns via the human-review channel or by emailing info.mimwise@gmail.com.
Data used to train AI
We do not use your account, profile, content or conversation history to train, fine-tune or evaluate AI models unless you have specifically opted in (see Section 7 of the Privacy Policy). The setting defaults to off and is reversible at any time. We follow EDPB Opinion 28/2024 and the UK ICO generative-AI guidance.
Children
MimWise is designed for learners aged 13 and over. Additional safeguards, privacy defaults and the parental-consent workflow for under-13s are documented in our Children's Privacy notice alongside the requirements of the UK ICO Age Appropriate Design Code.
Reporting concerns
If you encounter bias, inaccurate output, inappropriate content, or any other concern about our AI systems, contact us at info.mimwise@gmail.com. We take every report seriously and will investigate promptly.
Continuous improvement
Our risk register, sub-processor list and model lineage records are reviewed quarterly. Material changes to our AI systems trigger a re-assessment under Article 9 of the EU AI Act and an update to this page (with a refreshed “last updated” date).