1. Human flourishing
Technology should improve the capability, confidence and judgement of people within organisations. We will not treat workforce displacement as an automatic measure of progress.
2. Human authority and accountability
The Organisational Brain may advise, analyse, retrieve, generate and automate within agreed boundaries. Accountability for material decisions remains with identifiable people and the organisations that deploy the system.
3. Purpose limitation
AI capabilities should be used for clear, legitimate and documented purposes. Data, tools and permissions should be proportionate to the task, and secondary uses should be reviewed rather than assumed.
4. Evidence, uncertainty and explainability
Wirtor should distinguish sourced evidence from inference, opinion and generated content. Material outputs should expose relevant provenance, assumptions, limitations and uncertainty in a form suitable for the user and decision context.
5. Fairness and non-discrimination
Systems should be assessed for foreseeable unfair impact, including risks created by incomplete data, historical bias, proxy variables, inaccessible design or inappropriate use. Higher-impact uses require stronger review, testing and governance.
6. Privacy and data stewardship
Personal and confidential information should be minimised, protected and used only with appropriate authority. Access, retention and model interactions should be governed according to sensitivity and risk.
7. Safety, security and resilience
AI introduces risks including prompt injection, data leakage, manipulation, unsafe actions, unreliable outputs and dependency on third-party models. Wirtor will apply layered controls, scoped permissions, monitoring, testing and incident response proportionate to the potential harm.
8. Contestability and escalation
People affected by a material AI-supported process should have an appropriate route to question, correct or escalate an outcome. Systems should support intervention and safe suspension where a decision, data source or automated action is disputed.
9. Organisational context
Organisational knowledge is contextual. Wirtor should not collapse disagreement into false certainty or treat what is most frequently recorded as necessarily correct. Contradictions, minority evidence and unresolved questions can be valuable intelligence.
10. Proportionate governance
Controls should scale with impact. Experimental drafting assistance is not governed in the same way as a system influencing employment, safety, finance, legal rights or access to essential services. Consequential uses require documented ownership, approval, monitoring and review.
11. Continuous evaluation
AI performance and risk change over time. Wirtor will seek to review models, prompts, data flows, integrations and real-world outcomes, and will revise safeguards where evidence shows they are inadequate.
12. Transparency about limitations
We will not knowingly represent probabilistic outputs as guaranteed facts, hide material AI involvement, or claim that a control, certification or compliance status exists when it has not been established.