This checklist is designed for fast triage of a breaking AI story: enough to determine whether you’re looking at a tool malfunction, a systemic failure, or a governance story disguised as a tech story.

Each question is answerable from:

  • the article itself,
  • public statements,
  • or what is conspicuously not said.

A. Authority & Framing (Questions 1–3)

  1. Does the system present itself as giving “the answer” rather than assisting judgment?
    ☐ Present ☐ Absent ☐ Unclear
  2. Are users or institutions likely to defer to the output without cross-checking?
    ☐ Present ☐ Absent ☐ Unclear
  3. Is responsibility for errors framed as diffuse (“the AI,” “the data,” “users”) rather than owned by a specific actor?
    ☐ Present ☐ Absent ☐ Unclear

B. Interface & Information Loss (Questions 4–6)

  1. Is complex or high-stakes information summarized or compressed into a single output?
    ☐ Present ☐ Absent ☐ Unclear
  2. Are uncertainty, disagreement, or caveats missing from the default presentation?
    ☐ Present ☐ Absent ☐ Unclear
  3. Does the output risk reassuring users when follow-up, caution, or urgency would be appropriate?
    ☐ Present ☐ Absent ☐ Unclear

C. Differential Harm & Bias (Questions 7–8)

  1. Do harms or errors disproportionately affect specific groups (e.g. women, minors, patients, marginalized users)?
    ☐ Present ☐ Absent ☐ Unclear
  2. Does the harm arise through omission or framing rather than explicit instruction or prohibition?
    ☐ Present ☐ Absent ☐ Unclear

D. Governance & Aftermath (Questions 9–10)

  1. Are official responses limited to statements, pledges, or future fixes rather than immediate constraints?
    ☐ Present ☐ Absent ☐ Unclear
  2. Does the system continue operating at scale despite acknowledged problems?
    ☐ Present ☐ Absent ☐ Unclear

Rapid Read Guide (for newsroom use)

  • 0–2 Present → Likely a narrow technical issue
  • 3–5 Present → Product design or UX failure
  • 6–8 Present → Structural harm pattern
  • 9–10 Present → Governance failure story (not just “AI went wrong”)

Red flag combinations:

  • (1) + (4) + (6) → False authority causing quiet harm
  • (3) + (9) → Responsibility vacuum
  • (7) + (8) → Systemic bias via omission

What this checklist is optimized for

  • Writing a second-day story that moves beyond “experts warn…”
  • Identifying who should be asked to answer, not just what went wrong
  • Distinguishing:
    • bad output stories
    • from bad system stories
    • from bad governance stories