Transparency

Responsible AI

Last updated: October 2026

Summary

SmartRR uses simple automated checks to help our moderators spot reports that need a closer look. A person reviews every report. The system never decides what is true, and it never publishes anything.

What the automated checks do

Similar reports: flags text reports that are nearly identical but come from different numbers within 24 hours.
Bursts: flags an unusual number of reports for the same state and category in a short time.
Patterns: flags many reports from one number in an hour, or many numbers sending near-identical text.

A flag only adds a reason and moves the report toward the top of the review queue. USSD reports have no free text, so similarity checks apply to WhatsApp and app reports only.

What they do not do

They do not approve, reject, withhold or publish a report.
They do not identify, profile or score individuals.
They do not send anything to institutions on their own.

Human oversight

Moderators review every report. They can approve, correct and approve, route to an institution, or withhold, and must give a reason that is logged in an audit trail. Nothing appears on the public map without a moderator's approval.

Accuracy

Our target is fewer than 5% of flags being wrong (false positives). We check flag quality against historical data and monitor it weekly. A wrong flag never removes a report on its own.

Stop-use switch

The automated checks can be switched off at any time by [stop-use authority, TBD]. Reports continue to be reviewed by people.

Appeals and corrections

If you think your report was wrongly flagged or withheld, or want data corrected, email info@smartrrapp.org with your report reference. A different moderator will review the decision.