Responsible AI Statement

Last updated: 13 September 2026

AI Architect designs, builds and reviews AI-enabled software for organisations. We believe useful AI requires clear ownership, proportionate controls and honest communication about what a system can and cannot do.

Our principles

Purpose before technology

We start with the business need, affected people and operating context. AI is used only where it provides a defensible benefit over a simpler approach.

Human accountability

Organisations remain responsible for decisions made using their systems. We design appropriate human review, escalation and override points, particularly where errors could materially affect people, safety, rights or access to services.

Privacy and data control

We seek to minimise personal and confidential data, define approved data flows and assess how model and infrastructure providers store or use information. Client data should not be used to train third-party models unless that use is explicitly understood and approved.

Security by design

AI features are considered within the wider application threat model. Controls may include access management, input and output handling, secrets protection, logging, rate limits, retrieval boundaries and testing for misuse.

Fairness and accessibility

We consider who may be excluded, disadvantaged or disproportionately affected. The depth of testing and oversight should reflect the context and potential impact of the system.

Transparency

Users and operators should receive appropriate information about when AI is involved, the source and limitations of its outputs, and how to challenge or escalate an outcome.

Evaluation and monitoring

Demonstrations are not evidence of production readiness. We favour documented evaluation criteria, representative testing, operational monitoring and periodic review after launch.

Limits of AI output

Generative AI can produce incomplete, inaccurate or unexpected output. AI output should not be treated as professional advice or as the sole basis for a high-impact decision. Required safeguards depend on the use case, users, data and potential consequences.

Work we will question or decline

We may question, redesign or decline work where the intended use is unlawful, deceptive, unsafe, discriminatory, excessively intrusive or lacks credible accountability. This includes systems intended to manipulate people, bypass legitimate security controls or automate high-impact decisions without suitable review.

Shared responsibility

Responsible operation is shared between AI Architect, the client and relevant technology providers. Project-specific responsibilities, approved providers, testing, data handling and acceptance criteria should be recorded in the applicable agreement and Statement of Work.

Questions and concerns

Questions about this statement or concerns about an AI Architect project can be sent to info@ai-architect.uk.