Below is a binary, field-usable checklist derived directly from the 7 meta-themes, designed to be applied to any new AI incident, rollout, or controversy.

It is intentionally non-normative and non-ACP-specific.
Each item is Present / Absent / Unclear.
“Unclear” is important: it flags opacity as a finding, not a gap in analysis.

Meta-Theme 1: Unauthorized Epistemic Authority

1. The system presents outputs as settled answers rather than provisional inputs
   ☐ Present ☐ Absent ☐ Unclear
2. Users or institutions are likely to treat outputs as authoritative by default
   ☐ Present ☐ Absent ☐ Unclear
3. There is no explicit mandate defining where the system may or may not speak conclusively
   ☐ Present ☐ Absent ☐ Unclear
4. Responsibility for epistemic harm is diffuse or denied (“it’s just a tool”)
   ☐ Present ☐ Absent ☐ Unclear

Meta-Theme 2: Interface as Governance

5. The interface presents a single synthesized answer rather than multiple perspectives
   ☐ Present ☐ Absent ☐ Unclear
6. Uncertainty, disagreement, or evidence hierarchy is suppressed or hidden by default
   ☐ Present ☐ Absent ☐ Unclear
7. The design reduces friction for acceptance (e.g., confidence, placement, tone)
   ☐ Present ☐ Absent ☐ Unclear
8. Interface choices shape behavior without being framed as governance decisions
   ☐ Present ☐ Absent ☐ Unclear

Meta-Theme 3: Compression-Induced Harm

9. Complex or high-stakes information is summarized or compressed
   ☐ Present ☐ Absent ☐ Unclear
10. Caveats, edge cases, or context are omitted in the compressed output
    ☐ Present ☐ Absent ☐ Unclear
11. Outputs risk false reassurance or premature closure of inquiry
    ☐ Present ☐ Absent ☐ Unclear
12. Potential harm arises from omission rather than outright falsehood
    ☐ Present ☐ Absent ☐ Unclear

Meta-Theme 4: Bias as Structural and Uneven

13. Effects differ across populations (e.g., gender, age, vulnerability)
    ☐ Present ☐ Absent ☐ Unclear
14. Severity or urgency is downplayed for certain groups
    ☐ Present ☐ Absent ☐ Unclear
15. Bias manifests through framing, tone, or omission rather than explicit rules
    ☐ Present ☐ Absent ☐ Unclear
16. Bias appears model-specific or task-specific rather than universal
    ☐ Present ☐ Absent ☐ Unclear

Meta-Theme 5: Adoption Without Institutional Comprehension

17. Institutions rely on the system for triage, summarization, or workload reduction
    ☐ Present ☐ Absent ☐ Unclear
18. The institution cannot clearly identify model provenance, versioning, or changes
    ☐ Present ☐ Absent ☐ Unclear
19. Humans remain formally responsible but lack time or capacity to reconstruct omissions
    ☐ Present ☐ Absent ☐ Unclear
20. Operational reliance precedes formal understanding or auditability
    ☐ Present ☐ Absent ☐ Unclear

Meta-Theme 6: Governance as Performance, Not Enforcement

21. Public statements or pledges substitute for binding constraints
    ☐ Present ☐ Absent ☐ Unclear
22. Safeguards are announced without clear enforcement mechanisms
    ☐ Present ☐ Absent ☐ Unclear
23. Legal or regulatory responses are delayed relative to deployment
    ☐ Present ☐ Absent ☐ Unclear
24. Delay or review functions as de facto permission for continued operation
    ☐ Present ☐ Absent ☐ Unclear

Meta-Theme 7: Scale, Speed, and Irreversibility

25. The system deploys or propagates at large scale or high speed
    ☐ Present ☐ Absent ☐ Unclear
26. Correction or retraction mechanisms are slower than dissemination
    ☐ Present ☐ Absent ☐ Unclear
27. Harm persists after exposure, criticism, or partial remediation
    ☐ Present ☐ Absent ☐ Unclear
28. Downstream effects are difficult or impossible to fully reverse
    ☐ Present ☐ Absent ☐ Unclear

Quick Interpretation Guide (optional)

• 0–6 “Present” → localized or early-stage issue
• 7–14 “Present” → systemic risk emerging
• 15+ “Present” → de facto epistemic infrastructure failure

Clusters matter more than totals:

• Heavy presence in 1–2 → authority/interface problem
• Heavy presence in 3–4 → harm + bias problem
• Heavy presence in 5–7 → institutional/governance breakdown