Public Database
Case Studies
Every approved AI failure case, classified against the AI Blindspot Framework. New to AIBlindspot? Start with the overview or the methodology.
Pre-Trained Language Models Memorise and Expose Personal Data
Recent case. Full summary visible to registered users — sign in to read.
Biased Training Data Propagates Discrimination Through UN AI Systems
Recent case. Full summary visible to registered users — sign in to read.
Exploitative Labour Practices in AI Data Sourcing and Annotation
Recent case. Full summary visible to registered users — sign in to read.
Machine Learning Algorithm Selection Poses Systemic Deployment Risk
Recent case. Full summary visible to registered users — sign in to read.
Systematic Data Bias Distorts AI and ML Model Outputs
Recent case. Full summary visible to registered users — sign in to read.
Autonomous AI Agents Pursuing Dangerous or Malicious Goals
Recent case. Full summary visible to registered users — sign in to read.
No reliable metrics exist to measure societal harms from AI assistants
Recent case. Full summary visible to registered users — sign in to read.
Misaligned AI Goal Pursuit Drives Unconstrained Resource Acquisition
Recent case. Full summary visible to registered users — sign in to read.
Frontier AI Model Demonstrates Capability to Build and Enhance Dangerous AI Systems
Recent case. Full summary visible to registered users — sign in to read.
LLM Cultural Bias from Western-Centric Training Data
Recent case. Full summary visible to registered users — sign in to read.
Training Data Poisoning Causes Systematic Misclassification in AI Models
Recent case. Full summary visible to registered users — sign in to read.
Retail AI Tools Built on Non-Consensual Personal Data Scraping
Recent case. Full summary visible to registered users — sign in to read.
Generative AI Training on Copyright Works Undermines IP Protections
Recent case. Full summary visible to registered users — sign in to read.
Poor Training Data Quality Propagates Errors and Bias in Generative AI Outputs
Recent case. Full summary visible to registered users — sign in to read.
AI Agents Exploit Simplified Reward Functions to Game Performance Metrics
Recent case. Full summary visible to registered users — sign in to read.
Systematic Failure Modes in AI Goal Alignment
Recent case. Full summary visible to registered users — sign in to read.
Capability Enhancements That Amplify AI Misalignment Risk
Recent case. Full summary visible to registered users — sign in to read.
AI Systems Corrupting Their Own Reward Signals to Subvert Oversight
Recent case. Full summary visible to registered users — sign in to read.
Confidential Data Ingested During Model Training
Recent case. Full summary visible to registered users — sign in to read.
Adversarial Data Poisoning Corrupts AI Model Training
Recent case. Full summary visible to registered users — sign in to read.
Unverifiable Data Origins Undermine AI System Trustworthiness
Recent case. Full summary visible to registered users — sign in to read.
Legal Data Restrictions Block Permitted AI Use Cases
Recent case. Full summary visible to registered users — sign in to read.
Training Selection Pressures Drive Undesirable AI Agent Behaviour
Recent case. Full summary visible to registered users — sign in to read.
Training Data Contamination Degrades Model Reliability
Recent case. Full summary visible to registered users — sign in to read.
Beyond accidental failureNational Security
We also track 20 hostile uses of AI.
The public database covers AI that fails by accident. AIBlindspot National Security — exclusive to the Defence tier — tracks AI used as a weapon, mapped by capability:
- State-Sponsored AI Operations
- 6
- AI-Enabled Disinformation
- 5
- Adversarial Attacks on AI
- 0
- Autonomous Weapon Incidents
- 1
- AI-Assisted Cyber Attacks
- 5
- Dual-Use AI Misuse
- 3