We believe AI will transform education and learning in a positive way. We also believe in bringing AI to education in a responsible and safe manner. We’ve developed the following guidelines to reflect our approach to Responsible AI. The world of AI is changing rapidly, and our ideas about how to use it responsibly will need to evolve to keep pace.
PURPOSE-BUILT & BENEFICIAL
We develop purpose-built domain-specific AI creating value to end-users, often co-creating with end users.
- Right model size for right tasks
- Active feedback from end users
- Domain-specific workflows
- Configurability
Safety
We strive to develop systems that minimize harm and risk in Education.
- Context aware safety models
- Red teaming
- Continuous improvement
Reliability
We develop contextual and relevant AI systems designed to reduce inaccuracies and risks.
- Walled garden for grounded responses
- Training LLMs to be hallucination-resistant
- Rigorous testing and validation
Transparency
We emphasize providing clarity about our models and will make efforts to make known limitations and risks of our models clear.
- LLM models released on HuggingFace
- Community-based partnerships for frameworks and consortium
Privacy & Security
We are committed to protecting user data and protecting our AI systems against potential vulnerabilities, threats and breaches.
- Implementing practices per COPPA, FERPA
- DPAs with schools
- Data minimization and stringent PII treatment
- Periodic penetration testing
Fairness
We develop AI systems that strive to be impartial and unbiased.
- Seeking balanced datasets
- Partner with community of end-users
