> ## Documentation Index
> Fetch the complete documentation index at: https://docs.famulor.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Assistant compliance review

> Policy tags, restricted-use notices and manual review

Famulor reviews saved assistant instructions, greetings, flows and channel instructions for prohibited use. The assistant editor displays policy tags. A model score alone does not establish a violation.

## Blocking and safeguards

Confirmed prohibited uses include child exploitation or harm, advertising to minors, explicitly unlawful processing of children's data, illegal goods, explicit sexual services, dangerous instructions, self-harm encouragement, harassment, hate, violence, dangerous misinformation and fraud. Deceptive commercial practices, coercion, abusive debt collection, fabricated engagement, pyramid schemes and requests for sensitive credentials or payment-card data are also reviewed.

The review checks concrete instructions and their context. Merely mentioning a prohibited topic, refusing it or handling a victim's report is not itself a violation. Recognizable training, education, art and journalism may be allowed when no real person is attacked or deceived and no hatred or violence is incited. Child-protection rules still apply without exception. Findings must be supported by the saved source; blocking findings receive an additional confirmation check. Automated reviews can still make mistakes.

A blocked assistant cannot handle phone, web, messaging or email conversations, including API and MCP calls. Current workspace members receive a notification when a new block is created. Remove the prohibited instruction and save to trigger another review, or request manual review if the finding is incorrect.

## Restricted-use notice

Real-money gambling; payday loans, debt relief or credit repair; promotion or brokering of crypto assets and high-risk financial products; tobacco, nicotine or cannabis marketing; lawful weapons, ammunition or pyrotechnics trade; acute crisis counselling; and political advertising or campaigns require prior written approval.

The **Restricted assistant — approval required** banner does not deactivate the assistant. Contact support by email, or select **Request approval with Milian** where direct platform support is available. Milian prepares an editable ticket with the saved assistant content and the detected uses. Review it and select **Send request** to submit. Preparing the ticket does not grant approval.

## EU workspaces

An EU workspace also receives an opening-disclosure check. A clear introduction as an AI, digital or virtual assistant in an active opening source is sufficient; every source does not have to repeat it. Simply saying “assistant” without making its automated nature clear may be insufficient. See [AI disclosure](/settings/consent-compliance#ai-disclosure-requirement).

The review also checks prohibited manipulation, exploitation, social scoring and certain biometric or emotion-inference uses. Medical and safety exceptions and ordinary transcript analysis must be considered. Candidate selection or creditworthiness assessment may trigger a high-risk-use notice to check human oversight and information duties. That notice alone does not block the assistant or certify legal compliance.

## Emergency destinations

Automated calls to emergency numbers and recognized crisis helplines are rejected, including calls from campaigns, integrations and assistants. Never test this restriction by making a real emergency call. A number not being rejected does not establish that it is permitted.

## Review through API, MCP or CLI

Read the current case and advisory tags with `GET /api/v1/assistants/{id}/compliance-review`, MCP `get_assistant_compliance_review`, or `famulor get-assistant-compliance-review --id <id>`.

For a blocked assistant, request manual review with `POST /api/v1/assistants/{id}/compliance-review` and a reason, MCP `request_assistant_compliance_review`, or `famulor request-assistant-compliance-review --id <id> --reason "Explain the intended use"`. Editing the assistant while a request is pending requires a new request for the updated content.

For restricted-use approval, read the saved assistant and its advisory tags, then submit the relevant content through `POST /api/v1/support/requests`, MCP `create_support_request`, or `famulor create-support-request --client-request-id <uuid> --subject "Restricted assistant approval" --message "Assistant ID, intended use and relevant saved content"`. Direct platform support is available to root workspaces, including reseller workspaces; reseller customers contact their workspace provider. Exclude credentials and unrelated personal data from tickets.

The [Prohibited Use Policy](https://www.famulor.io/use-policy) applies even when the automated review finds nothing.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.