Scope
The ICU6 methodology covers citizen reports, public monitoring, and verified open-source sources.
- Fraud, identity impersonation, phishing, fake news
- Channels: email, SMS, social networks, web, messaging apps
- Areas: worldwide, with a focus on FR/EU and local extensions
Data sources
Data comes from a public form, partner tools, and compliant OSINT collections.
- Reports via ICU6.org (explicit consent)
- Browser extensions and plugins (opt-in)
- Open sources (institutional sites, CERT, trusted press)
- Partner feeds compliant with GDPR / contracts
Processing pipeline
Each alert follows a standardized, traceable, versioned pipeline.
- Ingestion & timestamping
- Cleaning, anonymization, and normalization
- AI analysis (categorization, risk signals, summary)
- Consecutive human reviews (2 levels)
- Publication to the public database / withdrawal if sensitive
AI analysis
We use audited AI models. The displayed scores are decision aids, not verdicts.
>= 90% Target precision
>= 85% Target recall
≤ €0.03 per analysis Cost ceiling (indicative)
Values are internal targets, reviewed through quality audits.
Human validation
Two volunteers review the AI analysis: an analyst, then an independent validator.
- Fixing classification errors
- Adding sources/evidence and risk indicators
- Decision to publish, strengthen anonymization, or withdraw
Privacy & compliance
Data minimization, strict anonymization, and limited retention are applied by default.
- No data sales; access is restricted and logged
- Deletion on request and short retention periods
- Compliance with GDPR / ePrivacy; DPIA for sensitive processing
Limits & biases
Despite our controls, some errors or biases may persist.
- Variable quality of reports (noise, duplicates)
- Fast-evolving threats (obsolescence)
- Uneven coverage by country and channel
Spotted an error?
Contact us to correct a record or request a takedown.
Versioning & traceability
Each alert keeps the history of analyses, decisions, and updates.
- Timestamps, authors, comments
- Version numbers and diffs
- Log of removals/re-activations
Contact us
For methodological questions, write to us via the contact form.
Response subject to volunteer availability.