Serious Concern Detection
Serious Concern Detection is an AI-powered safety feature that automatically flags feedback requiring immediate human review. It is designed for organizations where anonymous feedback may surface real harm — covering self-harm, threats, harassment, abuse, and severe misconduct.
It is especially important for organizations in education, healthcare, and other high-risk sectors where acting quickly on a concern can make a meaningful difference.
How it works
Every time a feedback response is submitted to a room, it goes through sentiment analysis. After that, the Serious Concern Detection system evaluates it:
- Scope check — by default, only negative-sentiment feedback is scanned. You can change this to scan all feedback regardless of sentiment (see Configuration).
- AI evaluation — the system evaluates the text using a layered prompt: core safety definitions, rules specific to your organization type, your chosen sensitivity level, and any custom criteria you've written.
- Flagging — if the AI determines the feedback meets the concern threshold, it is flagged and stored.
- Notification — team members with serious concern alerts enabled receive an email immediately.
- Resolution — a team member reviews the concern, takes any necessary action, and marks it resolved with an optional note.
Viewing flagged concerns
Go to Insights → Serious Concerns for any room. The default view shows only unresolved concerns — nothing slips through by default.
Each concern shows:
- The original feedback text
- The AI's reasoning for flagging it
- The event type (room) and instructor ID, if present
- A Jira ticket link if your org has Jira connected and a ticket was created
Use the Unresolved / Resolved / All toggle to switch views.
Resolving a concern
- Click Resolve on the concern card.
- Optionally add a resolution note (visible to other team members, useful for audit trails).
- Click Confirm.
Resolved concerns can be reopened at any time if follow-up is needed.
Who gets notified
Email alerts are sent per-user based on notification preferences. To receive serious concern emails:
- Go to Settings → Notifications
- Enable Serious concern alerts
Owners and admins are notified by default. Members can opt in. Notifications respect per-user preferences — team members who don't want alerts won't receive them.
Configuration
Go to Settings → Serious Concern Detection to configure how detection works for your organization.
Owners and System Admins always have full access to this settings page. Admins can access it only if the owner has enabled Allow admins to configure serious concern detection. Members cannot access it.
Feedback scope
| Setting | Behaviour |
|---|---|
| Negative feedback only (default) | Only feedback classified as negative sentiment is scanned. Recommended for most organizations. |
| All feedback | Every submission is scanned regardless of sentiment. Use in high-risk environments where concerns may be expressed neutrally or even positively (e.g. a student describing a situation using calm language). |
Detection sensitivity
| Level | Behaviour |
|---|---|
| Low | Only flags clear, explicit concerns. Minimizes false positives. Best for organizations where accuracy is prioritized over coverage. |
| Medium (default) | Balanced. Flags explicit and moderately ambiguous language. Recommended for most organizations. |
| High | Flags anything potentially concerning — including indirect or coded language. Best for healthcare, child services, or any environment where missing a concern is the greater risk. |
Custom criteria
Write natural language instructions describing what your organization specifically considers a serious concern. This is added on top of the base detection rules — it cannot override the core safety definitions.
- Up to 2000 characters
- Use bullet points for clarity
- Avoid overly broad criteria that would flag routine complaints
Six pre-built templates are available:
| Template | Best for |
|---|---|
| Education | Universities, schools — student mental health, inappropriate behavior, unsafe labs |
| Healthcare | Hospitals, clinics — patient safety, medication errors, staff misconduct |
| Enterprise | Large organizations — workplace harassment, discrimination, retaliation |
| SaaS | Software companies — fraud, data misuse, hostile behavior |
| Nonprofit | Volunteer and donor concerns, safeguarding issues |
| Government | Public sector misconduct, ethics violations |
Click a template to fill the criteria field, then edit it to match your organization's specific context.
Jira integration
If your organization has Jira connected, you can configure a Serious concern trigger that automatically creates a Jira issue whenever feedback is flagged. The ticket includes the feedback text, AI reasoning, and a link back to the concern in Anonfeedback.
Configure the Jira integration →
Recommended setup by sector
| Sector | Recommended scope | Recommended sensitivity | Custom criteria |
|---|---|---|---|
| K-12 / Higher education | All feedback | High | Yes — Education template |
| Healthcare | All feedback | High | Yes — Healthcare template |
| Enterprise / HR | Negative only | Medium | Yes — Enterprise template |
| SaaS / Tech | Negative only | Low | Optional |
| Nonprofit | Negative only | Medium | Yes — Nonprofit template |