Human decisions. Clearly identified AI.
ESH is designed around explicit AI disclosure, provenance labels and meaningful human control.
When you use an ESH AI assistant, the interface must identify that you are interacting with an AI system. AI outputs may be incorrect and are not a substitute for legal, tax, accounting, financial, investment or other regulated professional advice.
1. Where AI may be used
Potential uses include startup planning, educational assistance, search, recommendations, summarisation, translation, opportunity matching, content moderation support, market research assistance and creative-production tools.
2. Labelling standard
ESH can use labels such as AI-generated, AI-assisted, Company-provided information, Verified fact and Sponsored. These labels describe provenance and should not be used as quality endorsements.
3. Generated and manipulated content
Where applicable to an enabled AI system, ESH should implement the required machine-readable marking and user-facing transparency for generated or manipulated content. Deepfakes and other content subject to specific transparency duties require appropriate disclosure.
4. Recommendations and scoring
Funding matches, investor matches and startup-readiness indicators are decision-support tools. Their methodology and key factors should be explained. ESH must not claim that a match guarantees eligibility, investment or success.
5. Human review
High-impact editorial, verification, moderation, commercial and programme decisions should have defined human oversight. Users should have mechanisms to report material errors or misleading AI output.
6. Public-interest content
If AI-generated text on matters of public interest is published, ESH should apply the transparency and editorial-control treatment required by the applicable AI Act rules.
7. Governance
Before enabling a production AI provider, document the provider, data flows, retention settings, subprocessors, security, user notice, human oversight, output labelling and applicable AI-system role.