Responsible AI Principles
Responsible AI Principles
How DWRC intends to use AI responsibly, and where human authority must remain.
- Version
- 1.0
- Effective
- 16 September 2026
- Last updated
- 16 September 2026
- Status
- Establishment-stage policy — legal review pending
Scope of these principles
These are the principles DWRC holds itself to. They are forward-looking by necessity: DWRC operates no AI decision system today, and describing one would be false. They are stated now so that any future capability is built against a standard that already exists, rather than one written afterwards to justify it.
Human accountability
Responsibility does not transfer to a machine. Where a digital worker acts, accountability remains with the people and organizations who deploy, instruct, or rely on it. No DWRC process should let responsibility dissolve because automation was involved.
Provider neutrality
No AI provider, platform, or vendor is treated as authoritative because of who they are. A provider's records may be evidence; a provider does not thereby become the judge of a matter in which it has an interest. These principles are deliberately written without reference to any particular company.
Evidence before assumption
Conclusions should rest on evidence with an identifiable source and acknowledged limits — not on assumptions about what a system must have done, and not on the fluency or confidence of a generated answer.
Contestability
Anything consequential should be capable of being questioned. That means enough reasoning and record to make a challenge meaningful, and a route by which a person can actually raise one.
Proportionality
The rigour of a process should match what is at stake. A minor, clearly verifiable matter does not need the machinery appropriate to a serious one — and a serious one must never be handled with the machinery of a trivial one.
Non-discrimination and privacy
Automated analysis can reproduce and amplify bias present in data or design. DWRC treats that as a risk to be actively examined. Information should be collected and retained only as far as a legitimate purpose requires, with particular care where evidence contains personal or confidential material.
Transparency appropriate to context
Where AI assistance materially shapes an outcome, that should be disclosed rather than obscured. Transparency is not unlimited — confidentiality, privacy, and security can properly constrain what is published — but it should never be reduced to silence about whether automation was involved at all.
Human oversight and decision authority
Consequential determinations about people and organizations should remain subject to appropriate human and institutional oversight. AI assistance may organize evidence, surface inconsistencies, summarize, or translate; those are supporting acts, not authority. DWRC does not promise that every future process will involve identical human participation — that depends on what is at stake and on applicable law — but it does hold that a consequential outcome should never be a final, unexaminable automated output, and that a route to human review and correction should exist where it matters.
Contact
Questions about this document may be sent to [email protected].