AI Output Disclaimer

Nothing produced by a Trust Agent listing constitutes:

• Medical advice - LLM output is not a substitute for diagnosis, prescription, or treatment by a qualified clinician. If you face a medical emergency, call your local emergency number.
• Legal advice - LLM output is not a substitute for advice from a solicitor, barrister, attorney, or other qualified legal professional admitted in your jurisdiction.
• Financial advice - LLM output is not a substitute for advice from an FCA-authorised (or equivalent) financial advisor. Trust Agent is not authorised by the FCA and does not provide investment recommendations.
• Tax advice - Always consult a qualified accountant or tax advisor for jurisdiction-specific tax planning.
• Mental-health diagnosis or treatment - Companion or wellbeing roles are not therapists. If you are in crisis, contact your local crisis line (Samaritans 116 123 in the UK).

LLMs can confidently generate plausible-sounding statements that are factually wrong - this is sometimes called “hallucination”. It can affect names, dates, citations, code, statistics, and quoted text. The audit pipeline checks the listing’s prompt and behaviour spec for known unsafe patterns; it cannot verify every fact in every reply.

For citations, code, or any output you intend to publish or rely on, you must verify the content against an authoritative source.

LLMs reflect biases in their training data. Outputs may favour certain languages, cultures, perspectives, or demographic groups. The Trust Agent audit pipeline includes content-safety checks (SC-052 to SC-056) that flag overtly discriminatory language, but subtler bias may persist. Use AI-generated content with critical judgment, especially in hiring, lending, education, and other consequential decisions.

What you type into a session may be sent to a third-party LLM provider (Anthropic, OpenAI, Google, Mistral, Groq, or your own self-hosted model). Trust Agent does not log your message content, but the upstream provider may retain it under their own retention policy. Avoid sending non-public secrets, personal data of others without consent, or regulated data (health, payment, government identifiers) unless you have verified the upstream provider’s compliance posture.

Every audited listing carries a watermark line (visible at the bottom of role profiles) and a signed audit certificate at /audits/{id}. The certificate references the exact artefactHash that was reviewed. If a listing is updated after audit, a new audit must be completed before the new version can carry the badge.

Use the audit certificate to confirm the listing you are using is the one that was reviewed. A high trust score is a signal of audit rigour, not a guarantee of correctness.

If you bring your own LLM key or connect a local model (Ollama, LM Studio, Jan, llamafile), the Trust Agent audit applies only to the listing’s prompt and behaviour spec, not to the third-party model’s output. The third-party model’s safety posture is the responsibility of its operator (you, in the case of self-hosted models; the cloud provider in the case of BYO-key).

To the maximum extent permitted by law, AgentCore LTD is not liable for any loss or damage arising from your reliance on AI-generated content. This applies whether the loss is direct, indirect, consequential, financial, reputational, medical, or otherwise. Your remedies under this disclaimer are limited as set out in our Terms of Service.

If you encounter output that is illegal, unsafe, or in clear breach of our Acceptable Use Policy, report it to info@trust-agent.ai with the listing slug and a verbatim copy of the output. We will investigate and may revoke the listing’s audit certificate.

AgentCore LTD, 20 Wenlock Road, London, England, N1 7GU. Email: info@trust-agent.ai.