FWD-6 Synthetic content in claims and onboarding

Fijishi Working Draft 6 – 14 October 2026

Stage 1 of 4: working draft. Not yet open for public comment, not a numbered version, and not the criteria Fijishi certifies against. This draft replaces, as a new version, the earlier paper "Digital Authenticity Protocol" (2023-2026); the earlier text is available on request. Standards in development

The question

When documents, photos, video or voice submitted to the firm may have been generated or altered by AI, can the firm show how it checks them, and what it does when a check fails – without wrongly rejecting genuine customers?

Why it matters to firms

Claims evidence, identity documents and underwriting information can now be fabricated cheaply. Firms increasingly use AI to detect this – which is itself an AI decision about a customer. Both risks need an answer: accepting fabricated content, and rejecting genuine customers on the strength of an automated suspicion.

Principles

  1. Know where content is relied on: the points at which submitted content drives a decision.
  2. Check in proportion: checks match the value and risk of the decision.
  3. Use provenance where it exists: such as content credentials under the C2PA specification.
  4. A person decides a rejection: a failed or uncertain check leads to human review, not an automatic refusal.
  5. Record and monitor: checks and outcomes are recorded, and wrong flags are measured.

What a firm should be able to show (draft criteria)

  • FWD-6.1 A list of decision points where submitted content is relied on – for example claims evidence, identity and onboarding, and underwriting documents.
  • FWD-6.2 Defined checks for each point, including provenance signals where present.
  • FWD-6.3 Human review before any rejection based on a failed or uncertain check.
  • FWD-6.4 A record of each check and its outcome, as part of the decision's record (FWD-2).
  • FWD-6.5 Monitoring, on a sample, of how often genuine content is wrongly flagged.
  • FWD-6.6 A way for the customer to contest a decision based on a content check.

How it relates to existing rules and standards

  • It draws on the C2PA content-credentials specification where content carries them; it does not require them.
  • Identity verification, fraud and data protection rules apply on their own terms.

Where it is used now

It informs the Constitutional Baseline Review where a firm's system checks submitted content.

What happens next

Public comment follows FWD-1. Comments are welcome now: protocol@fijishi.com