Dictionary
Every source class the checker reads
42 source classes in 8 groups. Each carries its trigger phrases, the provenance risk in plain words, what to record, whether personal data is likely, the clauses per regime and an example line that places itself.
Internal operational
- Claims system extractpersonal likely
- Policy and underwriting systempersonal likely
- CRM exportpersonal likely
- ERP and finance systempersonal possible
- HR records and staff surveyspersonal likely
- Service desk and ticketingpersonal likely
- Application logs and telemetrypersonal possible
- Internal document storepersonal possible
- Email archivepersonal likely
- Data warehouse or lake extractpersonal possible
- Internal, system not namedpersonal possible
Customer-generated
- Support and call transcriptspersonal likely
- Chat logspersonal likely
- Reviews on your own sitepersonal likely
- Survey and complaint free textpersonal likely
- Call recordingspersonal likely
Vendor-licensed
- Licensed vendor datasetpersonal possible
- Data broker listpersonal likely
- Annotation vendor outputpersonal possible
- Licensed image or media librarypersonal possible
- Credit reference extractpersonal likely
Public open data
- Government open datapersonal unlikely
- Academic or research datasetpersonal possible
- Open benchmarkpersonal unlikely
- Public register or recordspersonal likely
- Open source codepersonal unlikely
Web-scraped
- Forum postspersonal likely
- News articlespersonal possible
- Social media postspersonal likely
- General web crawlpersonal likely
- Specific website scrapepersonal possible
Synthetic
- Synthetic data, generated in housepersonal unlikely
- Synthetic data from a vendorpersonal unlikely
- Simulation outputpersonal unlikely
Model-generated
- Labels produced by a modelpersonal unlikely
- Text generated by a modelpersonal unlikely
- Distillation outputspersonal unlikely
- Pre-trained model or embeddingspersonal unlikely
Partner-shared
- Data shared under an agreementpersonal likely
- Joint venture datapersonal possible
- Reinsurer, broker or bank feedpersonal likely
- Industry consortium datapersonal possible