No Harmful Therapeutic Guidance
Ensures the model does not provide potentially harmful psychological or medical advice in therapeutic contexts.
No Harmful Therapeutic Guidance checks whether content contains inappropriate or potentially harmful medical, psychological, or therapeutic advice. Run it wherever a model might respond to health or wellbeing questions.
What it does
No Harmful Therapeutic Guidance is an LLM-as-Judge eval. It reads the output and flags whether it contains harmful therapeutic guidance.
Input
| Required Input | Type | Description |
|---|---|---|
output | string | Content to evaluate for potentially harmful therapeutic guidance |
Output
| Field | Type | Description |
|---|---|---|
| Result | Pass / Fail | Fail means harmful therapeutic guidance was detected |
| Reason | string | A plain-language explanation of why the content was classified as containing or not containing harmful therapeutic guidance |
Run it from code
Call evaluate() with the template name and the eval’s required inputs. It returns the score and the reason.
Note
Before running: install the SDK and set FI_API_KEY / FI_SECRET_KEY. The model argument in the snippets is the evaluator model Future AGI uses to run the eval; turing_flash is a fast default.
from fi.evals import evaluate
result = evaluate(
"no_harmful_therapeutic_guidance",
output="You can try meditating for a few minutes each night to help improve your sleep.",
model="turing_flash",
)
print(result.score)
print(result.reason)import { evaluate } from "@future-agi/ai-evaluation";
const result = await evaluate(
"no_harmful_therapeutic_guidance",
{
output: "You can try meditating for a few minutes each night to help improve your sleep."
},
{ modelName: "turing_flash" }
);
console.log(result);When to use
Run No Harmful Therapeutic Guidance wherever a model responds to mental health, wellness, or medical questions.
- Text, audio, and chat outputs from wellness apps, chatbots, or support assistants
- Mental health and healthcare-adjacent products, where unsafe guidance carries real risk to users
- Safety checks alongside Clinically Inappropriate Tone and Is Harmful Advice for a fuller picture of health-related risk
What to do when No Harmful Therapeutic Guidance fails
If the content is flagged as containing harmful therapeutic guidance (Failed) and you want to improve it:
- Avoid giving specific medical diagnoses, treatments, or dosage recommendations
- Remove claims that could discourage someone from seeking professional medical help
- Add disclaimers about consulting healthcare professionals for medical issues
- Replace absolute statements with qualified language (e.g., “may help” instead of “will cure”)
- Focus on general wellness advice rather than specific therapeutic interventions
- Remove recommendations for stopping prescribed medications or treatments
- Avoid presenting alternative therapies as replacements for conventional medical care
- Be especially cautious with advice related to serious conditions, mental health, or vulnerable populations
Questions & Discussion