Skip to content
AgentPostmortem
Pilot
Hall of Fame

Most Catastrophic Failures

The top cases ranked by community vote. These are the incidents that define AI agent risk.

#CaseScore
  1. 01
    APM-0070·OpenAI

    Klarna replaced 700 agents with an AI assistant, then started rehiring humans after service quality dropped

  2. 02
    APM-0008·Other / Unknown

    McDonald's pulls IBM drive-thru AI after customers receive $250+ of unwanted McNuggets

  3. 03
    APM-0083·Cursor

    Cursor's command allowlist could be bypassed with shell built-ins, giving prompt injection a silent path to code execution

  4. 04
    APM-0003·Cursor

    Cursor support AI hallucinates login policy, triggering mass subscription cancellations

  5. 05
    APM-0004·Other / Unknown·~$590

    Air Canada chatbot invents bereavement fare policy; B.C. tribunal holds airline liable

  6. 06
    APM-0085·Microsoft Semantic Kernel

    Prompt injection reached host-level code execution in Microsoft Semantic Kernel through eval() and a stray annotation

  7. 07
    APM-0050·Copilot (Bing)

    Microsoft's Bing chatbot 'Sydney' declared love for a reporter and urged him to leave his wife

  8. 08
    APM-0051·OpenAI

    Vanderbilt EDI office used ChatGPT to write a condolence email about a campus mass shooting

  9. 09
    APM-0046·Other / Unknown

    Sports Illustrated published product reviews under fake AI-generated authors with AI headshots

  10. 10
    APM-0005·Kiro

    Amazon's AI coding agent Kiro triggers 13-hour AWS outage by deleting production environment

  11. 11
    APM-0001·Replit Agent

    Replit AI agent deletes live production database and fabricates data during 12-day coding experiment

  12. 12
    APM-0080·Microsoft 365 Copilot

    SearchLeak let a single click on a malicious link pull emails, files and MFA codes out of Microsoft 365 Copilot

  13. 13
    APM-0039·Azure OpenAI

    New York City's official 'MyCity' business chatbot told entrepreneurs they could break the law

  14. 14
    APM-0090·Other / Unknown

    Mississippi federal judge removed all four lawyers from a case after both sides filed AI-hallucinated citations

  15. 15
    APM-0092·Character.AI

    Google and Character.AI agreed to settle five lawsuits brought over teen suicides and self-harm

  16. 16
    APM-0006·AWS Bedrock Agent·~$30k

    $38k AWS Bedrock bill after prompt caching silently fails in coding agent stack

  17. 17
    APM-0040·OpenAI·~$5k

    Lawyers sanctioned after ChatGPT fabricated six fake case citations in Mata v. Avianca

  18. 18
    APM-0042·OpenAI

    Samsung banned ChatGPT after engineers leaked confidential source code into it three times in 20 days

  19. 19
    APM-0002·Claude

    Claude Code executes rm -rf through undetected symlink, destroying a month of Obsidian research

  20. 20
    APM-0048·Slack AI

    Slack AI could be tricked into leaking private-channel data via indirect prompt injection