AI Forensic Readiness v0.2 — Discussion Draft

The first three investigations exposed missing evidence structure; the remaining seven scenarios and incomplete validator/mapping backlog limited repeatability. This review build implements those areas while preserving the archived v0.1 work.

• Seven new reference bundles, completing the ten-scenario research backlog.
• Twenty deterministic failure/control conditions; supplemental models for 001-003.
• 158 committed normalized events and 120 manifested artifacts across the cases.
• v0.2 event profile, graph schema, explicit confidence, and missing-source limits.
• Consequence, chunk identity, memory lineage, parent/child grant lifecycle, approval presentation, and scoped validation evidence are all included.
• OCSF 1.6.0 and OTel Logs Data Model reference projections with original-envelope preservation.
• Full JSON Schema validation, conformance rejection fixtures, integrity tests, and CI.
• Case 003 review response, cross-case synthesis, practitioner briefing, and review kit are included.

Legacy v0.1 records remain valid under their original version. A separate v0.2 schema carries new requirements. Validation now requires the pinned dependencies; there is no silent fallback to partial validation. Public review remains open through October 17, 2026.

All cases are synthetic. Local state machines do not establish production control effectiveness, real-model behavior, independent reviewer performance, or live collector interoperability. OTel output is logical-model JSON, not OTLP wire JSON. Manifests provide consistency relative to trusted hashes, not signer authentication.

Specifications, explanatory text, and diagrams are licensed under CC BY 4.0. Schemas, examples, and code are licensed under Apache-2.0.