Air Canada
2024 · Documented AI Governance Failure
Their chatbot made a legal promise the airline then had to honor in court. No human in the loop. No escalation path.
UnitedHealthcare
2023 · Documented AI Governance Failure
An AI algorithm denied 90% of claims reviewed — more than any human reviewer had ever rejected. No override structure existed.
Zillow
2021 · Documented AI Governance Failure
An automated pricing model destroyed $500M in inventory. Leadership approved the AI strategy. Nobody owned the exit decision.
Amazon Rekognition
2020 · Documented AI Governance Failure
Facial recognition misidentified 28 members of Congress as criminals. The system was deployed before governance was built.
Apple Card
2019 · Documented AI Governance Failure
An algorithm issued credit limits 20× lower to women than men — including co-applicants at the same address. Structural bias, invisible at deployment.
Boeing 737 MAX
2019 · Documented AI Governance Failure
MCAS overrode pilots 346 times without their knowledge. No governance mechanism existed to intervene at execution.
COMPAS
2016 · Documented AI Governance Failure
A recidivism algorithm flagged Black defendants as high risk at twice the rate of white defendants. No audit structure caught it for years.
IBM Watson Health
2017 · Documented AI Governance Failure
Watson recommended unsafe cancer treatments that oncologists flagged immediately. No escalation path existed to stop deployment.
Optum Health
2019 · Documented AI Governance Failure
An AI tool systematically underserved Black patients by using cost as a proxy for need. The model ran for years before exposure.
Twitter/X
2020 · Documented AI Governance Failure
The photo-cropping algorithm consistently framed white faces over Black faces. Deployed globally before internal audit surfaced the pattern.
Air Canada
2024 · Documented AI Governance Failure
Their chatbot made a legal promise the airline then had to honor in court. No human in the loop. No escalation path.
UnitedHealthcare
2023 · Documented AI Governance Failure
An AI algorithm denied 90% of claims reviewed — more than any human reviewer had ever rejected. No override structure existed.
Zillow
2021 · Documented AI Governance Failure
An automated pricing model destroyed $500M in inventory. Leadership approved the AI strategy. Nobody owned the exit decision.
Amazon Rekognition
2020 · Documented AI Governance Failure
Facial recognition misidentified 28 members of Congress as criminals. The system was deployed before governance was built.
Apple Card
2019 · Documented AI Governance Failure
An algorithm issued credit limits 20× lower to women than men — including co-applicants at the same address. Structural bias, invisible at deployment.
Boeing 737 MAX
2019 · Documented AI Governance Failure
MCAS overrode pilots 346 times without their knowledge. No governance mechanism existed to intervene at execution.
COMPAS
2016 · Documented AI Governance Failure
A recidivism algorithm flagged Black defendants as high risk at twice the rate of white defendants. No audit structure caught it for years.
IBM Watson Health
2017 · Documented AI Governance Failure
Watson recommended unsafe cancer treatments that oncologists flagged immediately. No escalation path existed to stop deployment.
Optum Health
2019 · Documented AI Governance Failure
An AI tool systematically underserved Black patients by using cost as a proxy for need. The model ran for years before exposure.
Twitter/X
2020 · Documented AI Governance Failure
The photo-cropping algorithm consistently framed white faces over Black faces. Deployed globally before internal audit surfaced the pattern.