AI transparency

1. Our use of AI

AI-AO may use artificial intelligence as a support tool for research, drafting, translation, summarisation, software development, and image editing. AI supports our work; it does not replace our responsibility for what we publish or deliver.

2. Human review and editorial responsibility

Public website content and Knowledge Hub articles are reviewed by a person before publication. AI-AO retains editorial responsibility for published material. Important claims should be checked against suitable sources, and content may be corrected when new information becomes available.

3. When you interact with Open_H

Open_H is an authenticated generative AI assistant. Access requires acceptance of the current service information through a confirmation sent to the account email address. The interface identifies the assistant as AI and warns that its output may be inaccurate.

4. AI-generated and manipulated content

The chat interface identifies Open_H and its answers as AI. Where legally required, realistic synthetic or materially manipulated image, audio, or video content will be clearly disclosed. AI-generated text on matters of public interest will be disclosed when published without qualifying human review or editorial control. Machine-readable marking is preserved or added where supported by the configured model, output format, and publishing channel. Copying plain text can remove technical provenance, so anyone republishing an answer must retain a clear AI disclosure where required.

5. How Open_H produces answers

Open_H sends masked prompts and selected context to configured external AI and embedding providers. Retrieval-augmented generation searches administrator-approved document passages, and approved ontology relationships may add structured context. Citations identify retrieved sources, but a citation does not prove that the answer interpreted the source correctly.

6. Limitations and safeguards

AI systems can produce inaccurate, incomplete, biased, or outdated results. PII detection and masking reduce disclosure risk but cannot guarantee that every sensitive value is found. Users should not submit information they are not authorised to share and should not treat AI output as individual legal, medical, financial, security, or other professional advice.

Access, message frequency, and usage are limited. Administrators manage the knowledge base and may review masked technical and safety records. Open_H is not intended to make decisions that produce legal or similarly significant effects about people and must not be used as the sole basis for such decisions.

7. Roles, intended use, and AI literacy

Whether AI-AO, a customer, or another organisation is a provider or deployer under the EU AI Act depends on who develops, markets, configures, and uses a particular installation and for what purpose. Open_H is offered as a knowledge and drafting assistant; this page is not a declaration that every deployment has the same legal classification. Professional customers must assess their intended use before deployment, maintain appropriate human oversight and instructions, and ensure that staff operating or using AI on their behalf receive appropriate AI-literacy measures.

8. Prohibited and high-impact uses

Open_H must not be used for prohibited AI practices, including harmful manipulation or exploitation, social scoring, or unlawful biometric practices. Uses involving employment, education, essential services, credit, insurance, healthcare, law enforcement, migration, justice, or democratic processes require separate legal and risk assessment and must not proceed merely because the assistant is technically capable of producing an answer.

9. Questions, feedback, and legal information

Questions about our use of AI, reports of inaccurate or insufficiently labelled content, and suspected serious incidents can be sent to info@ai-ao.net. Further information is available in the EU AI Act and the European Commission's AI Act overview.