AI use notice

1. What DAWN is

DAWN is a conversational artificial intelligence system for recruitment, employment and workforce administration. It can explain information, retrieve authorised records, draft content, transcribe or generate speech, and prepare an action for review. It is designed to combine conversation with structured interfaces when confirmation, comparison or careful review is needed.

DAWN is not a person, lawyer, doctor, financial adviser or autonomous employer. It has no independent authority to hire, reject, promote, dismiss, discipline, set pay, approve leave or change a workforce record.

2. AI capabilities and providers

DAWN separates workforce policy from provider execution. Its AI orchestrator determines the permitted task, relevant policy, risk level, output structure and human-review requirement. Its LLM gateway routes an approved capability to a configured provider.

The configured primary provider may be replaced by an approved fallback when a request fails safely. Provider routing does not expand a user's access, alter the task's risk level or remove a human-review requirement.

3. Information used for an answer

DAWN may use the current message, permitted conversation history, selected attachments, relevant workforce records, the user's role and the policy needed to complete the request. It should use the minimum context reasonably needed for the selected task.

A user must not include information that is unnecessary, unlawful or outside their authority. Hidden instructions in an attachment or untrusted content must not override DAWN's access rules, system policy or approval requirements.

4. Human authority and meaningful review

Human review must be real, informed and performed by a person with appropriate competence and authority. The reviewer must have enough time and information to:

• understand the purpose and limits of the AI output;
• check the relevant source information and material assumptions;
• identify obvious errors, missing context and possible bias;
• seek further evidence or the affected person's view;
• correct, override or disregard the output; and
• stop the proposed action without an adverse consequence for doing so.

A person who merely confirms an AI recommendation without this independent assessment does not provide meaningful human oversight.

5. Recruitment, employment and other high-impact uses

AI used to screen, rank, match, assess or recommend decisions about candidates or workers can materially affect fundamental rights and may be regulated as high-risk AI depending on its intended purpose and deployment. The responsible organisation must assess the use case before deployment and document its role and obligations under applicable data protection, equality, employment and AI law.

Where required, the organisation must complete a data protection impact assessment and any applicable fundamental-rights assessment, define human oversight, maintain appropriate logs, monitor performance and bias, train affected staff, inform workers or their representatives, and provide affected people with a way to seek an explanation and contest an outcome.

Interview analysis generated by DAWN is marked for human review. It must not be treated as a verified score or propagated into a final candidate outcome until an authorised reviewer has assessed it.

6. Prohibited and unacceptable uses

DAWN must not be used to:

• infer a person's emotions in the workplace or during recruitment from biometric data, except where a strictly applicable medical or safety exception is lawful;
• categorise people from biometric data to infer race, political opinions, trade union membership, religion, beliefs, sex life or sexual orientation;
• make or implement a solely automated decision with legal or similarly significant effect unless a valid legal exception and all required safeguards apply;
• discriminate unlawfully or use protected characteristics, obvious proxies or irrelevant personal information to disadvantage a person;
• manipulate or deceive a person in a way that materially impairs informed choice or is likely to cause significant harm;
• create workplace social scores or penalise people based on unrelated behaviour or predicted personality;
• fabricate evidence, impersonate a person, conceal material AI use or misrepresent AI output as verified human judgement; or
• bypass permissions, monitoring, notice, consent, review or accountability controls.

DAWN may refuse, limit or require additional approval for a request that conflicts with these rules.

7. Voice, telephone and interview interactions

DAWN can convert speech to text, generate spoken responses and initiate configured telephone calls. A person must be told that they are interacting with AI where required by law or where it is not otherwise obvious. Recording and transcription require a separate lawful basis and any notice or consent required in the participant's location.

A voice, accent, pace, disability, hesitation or background sound must not be treated as reliable evidence of competence, truthfulness, emotion, health or personality. A transcription can omit or misinterpret words and must be checked before it affects an employment record or decision.

8. Accuracy, bias, robustness and monitoring

Generative AI can misunderstand intent, omit facts, reproduce historical bias, rely on unsuitable data, fabricate content or return an answer that is no longer current. A confident answer is not proof of accuracy.

Before production use, the responsible organisation must test the configured use case with representative data and foreseeable edge cases. Monitoring should consider accuracy, error rates, accessibility, differential outcomes, provider changes, fallback behaviour and user complaints. Material degradation must lead to restriction, correction or suspension of the affected capability.

Legal, health, safety, pay, tax, immigration and employment-law information must be verified by a suitably qualified person using an authoritative source.

9. Transparency to candidates, workers and users

Affected people must receive clear information about material AI use before it affects them. The responsible organisation must explain the purpose, the categories of data used, the role of the AI, the role of the human reviewer, the expected consequence, the available review route and the relevant privacy contact.

Where law requires it, workers and their representatives must be informed before a high-risk workplace AI system is put into use. Information must be accessible and understandable to the people affected, not limited to technical documentation for administrators.

10. Your controls, concerns and changes

You can ask DAWN to clarify a response, correct the information you supplied, stop a proposed action or retain a draft without sending it. You can ask the responsible organisation whether AI contributed to an outcome, request meaningful human review, express your view and contest a qualifying decision.

Report harmful, biased, misleading or unexpected behaviour through DAWN Support. Include the feature, date and time, what you expected, what occurred and the trace identifier if one is shown. Do not include unnecessary sensitive data.

We update this notice when DAWN's material AI capabilities, safeguards or legal responsibilities change.