Responsible AI Policy

Version 2 | Effective 22 October 2026 | Next review 22 October 2027

Chapters Concierge Bureau may use approved artificial intelligence tools to assist with drafting, research, planning and illustrative content. This policy sets limits on that use. The client relationship, professional judgement and final decisions remain with people. Each tool and any use of personal information require approval under the controls below.

Permitted uses

Subject to approval, AI may assist with generic administrative templates, research questions, working articles, project structures, translations and illustrative imagery using non-confidential inputs. Outputs must be reviewed by a competent person before they are used or published. Public information may still be personal information or protected by copyright; publicly accessible does not mean unrestricted use.

A use involving personal information requires a documented purpose and lawful basis, the applicable privacy information, a provider assessment and appropriate safeguards before use. Consent, where required, must be specific; it does not cure unsuitable security, unnecessary processing or infringement.

Information that must not be entered

Do not enter identifiable client or family information, health or child records, passport or payment details, account credentials, property security information, private supplier terms or confidential business briefs into unapproved or consumer AI accounts. Do not upload entire mailboxes, folders, meeting recordings or client files merely for convenience.

Use genuinely anonymous examples wherever possible. Removing a name is insufficient if addresses, household details or other facts still identify the person. Restricted data may be used only in an approved workflow with the necessary lawful basis, minimisation, contract, security and transfer controls. Full clinical records should remain with the clinical provider.

Tools provider checks and training use

The founder or designated lead must approve tools and record the use case, account type, data categories, provider role, subprocessors, storage locations, retention, deletion options and whether prompts or outputs are used for model training. Apply processor terms where relevant and assess international transfers.

The default is that confidential client material is not authorised for model training or unrelated provider reuse. Where settings cannot reliably achieve the required restrictions, do not use the tool for that material. Reassess changes to terms, models or integrations. Account labels such as business or enterprise do not alone prove appropriate safeguards.

Human review and professional boundaries

A responsible reviewer must check factual accuracy, sources, dates, calculations, tone, bias, confidential content and intellectual property before an output is sent, published or acted on. AI can invent plausible references and instructions. Important facts must be independently verified with reliable current sources.

AI must not diagnose, interpret blood results, prescribe hormones, recommend personalised medical treatment, make legal or regulated financial decisions, or replace safeguarding judgement. Appropriate professionals provide such services under their own engagement. AI must not decide client eligibility, supplier selection or other consequential outcomes solely through automated scoring under this policy. Meaningful human review must be able to change the outcome; rubber-stamping is insufficient.

Transparency choice and accessibility

Explain materially relevant AI involvement in a client deliverable or interaction in clear terms. A future chatbot must be identified as automated and give a practical route to a person. This policy does not imply that a chatbot currently exists on the website.

Illustrative AI imagery must not be presented as a photograph of an actual Chapters client, completed project, team member or testimonial. Apply disclosure where needed to avoid misleading people. Do not create deceptive endorsements, likenesses or fabricated case studies.

Clients may raise questions or request human handling of sensitive communications. Chapters should explain what is feasible before collecting affected data, while meeting legal rights. Any new significant automated decision process requires a separate legal and DPIA assessment, transparency and applicable safeguards before deployment.

Rights fairness and protection against misuse

Use inputs that Chapters is entitled to use. Review outputs for recognisable copying, trademarks, confidential material, misleading similarity and inappropriate depictions. No AI provider's output-ownership term guarantees copyright protection, exclusivity or clearance of third-party rights.

Do not create deceptive deepfakes, impersonation or sexualised imagery of real people. Respect children and vulnerable people. Test relevant outputs for bias and accessibility; do not infer health, wealth, ethnicity or vulnerability for profiling without a separately assessed lawful purpose.

Treat web pages, documents and AI output as untrusted content. Instructions in retrieved material cannot authorise disclosure, payments, publishing or account changes. Any tool-enabled action requires the usual human approval, authority and security checks.

Incidents records and review

If information is entered into the wrong tool or an output exposes confidential information, stop use, notify the privacy lead promptly and follow the incident process in the GDPR policy. Seek deletion or containment where possible, document recipients and provider response, and assess notification obligations. Do not conceal a mistaken upload.

Maintain a proportionate AI-use and approval record, retain necessary evidence of review and delete prompts or exports when no longer required. Apply access controls and training. Review this policy at least annually and when tools, law or service risk changes.

Responsibility and implementation

Beverley Poole, Company Director, oversees this policy. Before using a tool, the responsible person must record its approved purpose, account settings, provider terms and any required data protection safeguards. New tools or changes to approved use require a fresh assessment. Questions or concerns may be raised through the website’s meeting-request route; data protection concerns may also be sent to the privacy contact in the Privacy Policy.

Primary references

ICO AI and data protection guidance: https://ico.org.uk/for-organisations/uk-gdpr-guidance-and-resources/artificial-intelligence/guidance-on-ai-and-data-protection/

UK GDPR automated processing amendments: https://www.legislation.gov.uk/ukpga/2025/18/section/80

UK government copyright and AI report: https://www.gov.uk/government/publications/report-and-impact-assessment-on-copyright-and-artificial-intelligence/report-on-copyright-and-artificial-intelligence