01.
Human accountability
A person is responsible for every consequential decision an automated system supports. Systems assist judgement; they do not replace responsibility.
Responsible AIDIYAN AI / HYD
Commitments that shape our engineering decisions
Responsible practice is an engineering constraint at DIYAN AI, applied from design onwards. These are the standards our engineering work is held to.
Our commitments
Six standards we apply consistently
02.
Transparency about capability
We describe what a system can and cannot do in plain language, including its known limitations and the conditions it was evaluated under.
03.
Data minimisation
Collect the least data a system needs, keep it for the shortest reasonable period and restrict access to those who require it.
04.
Fairness in evaluation
Evaluation should cover the range of people and situations a system is intended to serve, with results reviewed before any deployment decision.
05.
Safety by design
Guardrails, fallbacks and escalation paths are part of the initial design, not additions after an incident.
06.
Continuous review
Our approach calls for ongoing monitoring and reassessment as operating conditions change. Behaviour that drifts outside expectations should be corrected or withdrawn.
In practice
Review is part of shipping
Our engineering approach calls for assessment of intended use, data handling, failure behaviour and escalation paths before a system reaches users. Systems that cannot meet those requirements should not ship.