AI ethics & responsible AI

Last updated 21 July 2026

Nexiour builds AI systems for other organisations, so how we build them matters. These principles describe how we approach responsible AI in practice — human oversight, data governance, evaluation and transparency — as commitments we hold ourselves to on every engagement.

Human oversight

We design AI systems to keep a human in control of consequential decisions. Agents and models augment people's judgement rather than replace accountability for it, and high-impact actions are built to be reviewable and reversible.

Data governance

We use client and personal data only for the purpose it was provided, under the terms agreed for the engagement. We do not use client data to train third-party foundation models without explicit permission, and we minimise the data a system needs to do its job.

Evaluation and fairness

We build AI systems behind real evaluation rather than a demo, and we test for failure modes — including biased or harmful outputs — appropriate to the system's use. Where a system affects people, we treat fairness and error analysis as part of the definition of done.

Transparency

We are clear with clients about what a system can and cannot do, its limitations and its failure modes. We favour designs that can explain or cite their outputs, such as retrieval-backed answers, over opaque ones where the use case allows.

Security and privacy

Responsible AI depends on secure AI. We apply the practices set out on our security page to the systems we build, and we treat privacy as a design constraint, not an afterthought.

Accountability

These are commitments, not a compliance certificate. We hold ourselves to them, we welcome being held to them, and we will update this page as our practice and the surrounding regulation evolve.

Questions about this document? Reach us through the contact page.