Risk-tiered AI development that classifies the system honestly, then builds in the transparency, human oversight and documentation that classification demands.
Obligations follow the risk a system creates for the people it affects, not the sophistication of the model behind it.
What it covers
Regulation (EU) 2024/1689 classifies AI systems by risk (unacceptable, high, limited and minimal) and attaches obligations to each tier covering data governance, technical documentation, transparency, human oversight, accuracy and robustness. Alongside it we apply established ethical AI principles: fairness, explicability, human agency and accountability.
How it changes delivery
Intended use is classified before development begins, prohibited practices are ruled out outright, and the duties attached to the relevant tier become product requirements rather than paperwork. Fairness and explicability are treated as design constraints that shape the interface, not disclaimers added to it.
Applied through
- Risk classification agreed before development begins
- Prohibited practices ruled out at the design stage
- Clear disclosure when a user is interacting with an AI system
- Human oversight, override and fallback designed into the flow
- Technical documentation, data governance and logging maintained