On Health Loops.
It answers, it does not diagnose. In the instructions and in the copy the member sees.
Every answer names its source. An answer without a source is a guess.
A person is one step away. Red flags, self-harm language and pregnancy questions go to a human.
The member controls the data. Consent at the first message, a way to delete, POPIA and GDPR duties met.
No training on member data. In the contract and in the configuration.
Questions
What are guardrails for health AI agents?
Guardrails are the rules a health AI agent must keep: what it must never do, what it must hand to a person, and what it must say instead.
Can an agent make health claims about our products?
No. It answers from your own labels and guides, and claim guardrails check supplement and medical wording before a reply goes out.
Can sensitive actions wait for a person?
Yes. A refund over your limit, for example, waits for someone on your team to approve it.