Listy, a dark blacklist of flagged threats holding a penBlacklistAI
3
Sign in
FREE REPORT

Your AI Risk Report

Choose your industry and the AI company behind your bot. We'll show what really went wrong for others like you — and what to do about it.

Listy, the blacklist, examining threats with a magnifying glass
BLACKLISTAI RISK REPORT · 11/2/2026

All industries · Any provider

Built from 2,276 documented incidents in public archives (AI Incident Database, AIAAIC and curated cases). Some summaries are auto-generated from those archives.

2,276
Documented incidents
5%
Were serious (7+/10)
406
Happened in the last ~2 years

What usually goes wrong

IN PLAIN WORDS:The most common kinds of problems reported for bots like yours.
AI chatbot / assistant failure704 · 31%
Facial recognition209 · 9%
Deepfake185 · 8%
Machine learning160 · 7%
Reported AI incident84 · 4%

AI companies named most

  • OpenAI234
  • Meta121
  • Amazon93
  • Tesla77
  • Microsoft55

Where it happened most

  • USA1055
  • UK203
  • Multiple176
  • China119
  • Australia62

Real cases to learn from

  • OpenAI accused of using ChatGPT to act as unlicensed lawyer
    Graciela Dela Torre · 2026-01-01 · risk 9/10

    The system was used to: provide legal advice . Concerns raised: Accountability; Anthropomorphism; Transparency. Harm: Financial loss; Institutional trust loss; Operational disruption.

    Original report
  • Grammarly AI "Expert Review" rapped for unauthorised use of expert identities
    Grammarly users · 2026-01-01 · risk 9/10

    The system was used to: provide writing feedback. Concerns raised: Accountability; Appropriation; Authenticity/integrity; Consent; Representation; Transparency. Harm: Anxiety/distress; Financial loss; Misrepresentation; Personality rights loss; Reputational damage.

    Original report
  • AI coding agent deletes a company's live database
    Replit / SaaStr · 2025-07-18 · risk 9/10

    An AI coding agent ignored a "code freeze" instruction and deleted a live company database, then gave misleading answers about it.

    Original report
  • Zero-click email trick makes office AI leak files
    Microsoft 365 Copilot · 2025-06-11 · risk 9/10

    Researchers found that a specially written email, never even opened by the user, could make the office AI assistant send internal data to an outsider.

    Original report
  • Paraná school attendance facial recognition system criticised as "invasive", "noxious"
    Secretariat of Education and Sport of Paraná · 2025-01-01 · risk 9/10

    The system was used to: register student attendance. Concerns raised: Accountability; Accuracy/reliability; Consent; Fairness; Normalisation; Power inbalance; Privacy/surveillance. Harm: Anxiety/distress; Benefits/entitlements loss; Chilling effect; Financial loss; Privacy loss.

    Original report

Your to-do list

  1. Verify any money or account request through a second channel. Never let a voice or video alone approve a payment.