Thanks for contributing to the Centre for AI Security and Integrity research repository.
- Regulatory mapping improvements (YAML/JSON updates with sources)
- Detection and classification signatures
- Methodology clarifications and reproducibility fixes
- Anonymized findings that fit existing schema contracts
- Reproducibility is required for all claims.
- Every headline number must map to an artifact and query in
claims/. - Use neutral research tone in report text; avoid promotional language and product CTA copy.
- Data submissions must be anonymized and schema-valid.
- Include source citations for regulatory/control mappings.
- Update or add schema docs in
schemas/if contract changes are introduced. - Add or update claim-ledger entries in
claims/when metrics change. - Add reproduction commands in report methodology docs.
- Update preregistration and citation logs when claim scope or legal references change.
- Ensure report templates/manuscripts include:
- headline integrity block (headline, denominator, run ID, artifact, query)
- fixed disclosure headings:
Limitations,Threats to Validity,Residual Risk,Reproducibility Notes
- Run relevant validation scripts in
pipelines/*/validate.sh.