Method
Building evidence that can be checked
The methodology is being designed for scrutiny: what a record should contain, what automation may and may not do, where human judgment enters, and how the resulting evidence can be independently tested.
What a record is
A Civic Shield record is designed to preserve more than a screenshot. The proposed evidence schema includes the source address, capture time, the page as served, relevant text extracted from images, and original-language content alongside translation. Preserved material can be fingerprinted using SHA-256 so that subsequent alteration becomes detectable. Access, edits and exports can be logged to create an auditable history of how the record was handled. The objective is simple: evidence should be capable of being examined by someone who did not collect it and who does not have to trust Civic Shield in order to test it.
Before Civic Shield publishes assessments of antisemitic content, the framework used for those assessments will be identified and published, together with the principles used to address context, ambiguity and borderline cases.
Political disagreement is not a proxy for antisemitism. Criticism of Israel, Israeli governments or their policies — and support for, criticism of, or opposition to Zionism — is not, by itself, treated as antisemitic.
Non-negotiable boundaries
- We do not name or publicly label individuals as antisemitic.
- We do not build profiles of individuals.
- We do not reproduce harmful content. Documentation should preserve evidence without becoming another channel for its distribution.
- We do not read private messages, and we do not enter closed spaces.
- We do not organise reporting swarms or automated pressure campaigns.
- We do not publish blacklists.
- We do not classify political positions as antisemitic simply because they concern Israel or Zionism. Criticism of Israel, Israeli governments or their policies — and support for, criticism of, or opposition to Zionism — is not, by itself, treated as antisemitic. Assessment depends on content and context under Civic Shield’s published framework.
- An automated system does not independently send notices, complaints, legal filings or other external escalations. Any such action requires human review and approval.
- Where collection is automated, it is read-only. It does not post, reply, engage or mass-report.
Testing platform accountability
Reporting harmful content asks an institution to act; measuring what happens afterwards asks whether the system actually responds. That measurement is a capability Civic Shield intends to test, and it is set out on platform accountability.
Status
The seal
Trust is useful. Verifiability is better. This demonstration shows one basic principle of evidence integrity: alter the illustrative record and its digital fingerprint changes. Restore it exactly and the original fingerprint returns.
Illustrative record fragment — not Civic Shield evidence
4
SHA-256 of the fragment
a86fe4bc75e50c22eb8ed043b9a9b084df8dfc305e6159a89a8c459a4d7f08e7
Illustrative seal generated for this demonstration · not yet verified in this browser
Recomputed locally in your browser. The demonstration does not require the record to be sent to a server.
Questions
Scrutiny is part of the method. If you work in research, law, policy, trust and safety, civil society or philanthropy and want to test the assumptions behind Civic Shield, we want that conversation. Write to us.