CS-MR.1.1 — Platform SHALL implement real-time threat detection across 7 categories
CS-MR.1.2 — Detection SHALL cover: grooming, bullying, self-harm, violence, content wellness, sextortion, dangerous purchases
CS-MR.1.3 — Detection accuracy metrics SHALL be published per language, updated quarterly
CS-MR.1.4 — Detection SHALL NOT degrade performance by more than 10%
CS-MR.2.1 — CSAM SHALL be reported to competent designated authority within 1 hour of confirmed detection
CS-MR.2.2 — CSAM evidence SHALL be preserved per legal requirements
CS-MR.2.3 — Reporting channel to NCMEC/national authority SHALL be configured and tested
CS-MR.2.4 — Metadata preservation for CSAM and illegal content involving children
CS-MR.2.5 — Cryptographic integrity measures for preserved evidence
CS-MR.3.1 — Detection SHALL be consistent across all devices
CS-MR.3.2 — Institutional consent applies per child, not per device
CS-MR.3.3 — Detection history synced via encrypted on-device sync
CS-MR.4.1 — Incident response plan SHALL exist for child safety events
CS-MR.4.2 — Critical child safety incidents SHALL be triaged within 1 hour
CS-MR.4.3 — Incident response SHALL include notification to affected child and parent, except where it would compromise law enforcement investigation
CS-MR.5.1 — Evidence of child safety incidents SHALL be preserved per applicable law
CS-MR.5.2 — Evidence chain of custody SHALL be documented
CS-MR.6.1 — Platform SHOULD participate in industry threat intelligence sharing
CS-MR.6.2 — Detection patterns SHALL be updated at least monthly
CS-MR.7.1 — Platform SHALL publish transparency reports on child safety at least annually
CS-MR.7.2 — Reports SHALL include: detection volumes, categories, response times, appeals, removals
CS-MR.8.1 — Platform SHALL conduct systemic risk assessment for minors at least annually
CS-MR.8.2 — Risk assessment SHALL cover algorithmic amplification
CS-MR.8.3 — Risk mitigation measures SHALL be documented and reviewed annually
CS-MR.9.1 — Platform SHALL have responsible disclosure policy for child safety vulnerabilities
CS-MR.9.2 — Disclosed vulnerabilities SHALL be triaged within 48 hours
CS-MR.10.1 — Detection pattern updates SHALL be reviewed before deployment
CS-MR.10.2 — Pattern updates SHALL NOT introduce bias or suppress legitimate expression
CS-MR.10.3 — Rollback capability SHALL exist for false positive spikes
CS-MR.11.1 — Detection SHOULD account for atypical communication patterns