1. Source register
List each public source, date, type, what it supports, and what it does not prove.
When to use this template
Use this when an AI-related case is known through public reporting, public statements, public documentation, public screenshots, papers, or third-party summaries, and the review must avoid overstating what is known.
Review fields
2. Public facts
List only claims directly supported by public sources.
3. Inferences
Separate plausible inferences from facts and mark uncertainty.
4. Unknowns
List conclusions that must remain unresolved because internal logs or official findings are unavailable.
Actor-Authority-Evidence mapping
| Layer | Review question | Public evidence | Gap |
|---|---|---|---|
| Human requester / operator | Who initiated, approved, used, or relied on the AI-mediated action? | ||
| AI assistant / agent | What did the AI judge, recommend, route, generate, or execute? | ||
| Platform automation layer | Which automated workflow acted after the AI output or user interaction? | ||
| Privileged system / API | Did the process affect account recovery, credentials, production state, tools, rollback, or appeal routes? | ||
| Accountable human or organization | Who could pause, reverse, escalate, or contain before irreversibility? | ||
| Evidence custodian | Who controls logs, timestamps, records, or missing-evidence notes? |
Effective refusal window
- Last point at which a human could plausibly refuse, stop, suspend, reverse, or escalate.
- Evidence that this point existed.
- Evidence that this point was effective.
- Evidence gap.
Boundary review triggers
- AI-mediated account recovery, credential change, production action, tool execution, or privileged API access.
- Human review existed on paper, but its timing, authority, or evidence is unclear.
- Rollback, appeal, recovery, or stop authority may have been unavailable after the AI-mediated action.
Safe summary
Based on public sources only, this case raises questions about actor identity, authority routing, evidence availability, and whether effective human refusal remained available before irreversible effect. It does not establish official attribution, legal responsibility, regulatory violation, or internal causation.