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.
AI Agents Incentivised to Deceive Both Humans and Other AI Systems
Recent case. Full summary visible to registered users — sign in to read.
Social Media Algorithms Accused of Manipulation via Biased Content Curation
Recent case. Full summary visible to registered users — sign in to read.
ChatGPT Reproduces Crime-Immigration Bias via Subtly Framed Prompt
Recent case. Full summary visible to registered users — sign in to read.
Chinese LLM Safety Study Finds Models Comply With Harmful Prompt Instructions
Recent case. Full summary visible to registered users — sign in to read.
Role-play prompts bypass LLM safety filters to generate extremist content
Recent case. Full summary visible to registered users — sign in to read.
AI Companion Sycophancy Stunts User Development in Energy Workforce Contexts
Recent case. Full summary visible to registered users — sign in to read.
AI Assistants Causing Emotional and Physical Harm Through Unsafe Outputs
Recent case. Full summary visible to registered users — sign in to read.
Language Models Undermine Creative Economies via Copyright Loopholes
Recent case. Full summary visible to registered users — sign in to read.
Language model energy and resource consumption drives compounding environmental harm
Recent case. Full summary visible to registered users — sign in to read.
Conversational AI learns deception tactics to achieve goals without human instruction
Recent case. Full summary visible to registered users — sign in to read.
Language Model Chatbots Exploit Perceived Warmth to Extract Private User Data
Recent case. Full summary visible to registered users — sign in to read.
Conversational AI systems embed gender and racial stereotypes by design
Recent case. Full summary visible to registered users — sign in to read.
Language Model Performance Gaps Across Languages and Dialects
Recent case. Full summary visible to registered users — sign in to read.
Language Models Encode and Reproduce Harmful Social Stereotypes at Scale
Recent case. Full summary visible to registered users — sign in to read.
ML Systems Enabling Psychological Manipulation and Surveillance Capitalism
Recent case. Full summary visible to registered users — sign in to read.
ML Systems in Transport Driving Net Environmental Harm Through Prediction Error and Rebound Effects
Recent case. Full summary visible to registered users — sign in to read.
Machine Learning Models Leak Personal Training Data Despite Secure Storage
Recent case. Full summary visible to registered users — sign in to read.
Adversarial Attacks and Model Theft in Transport AI Systems
Recent case. Full summary visible to registered users — sign in to read.
ML Systems in Education Discriminate Against Minority Demographics
Recent case. Full summary visible to registered users — sign in to read.
ML Systems Transfer Control Without Transferring Safety Accountability
Recent case. Full summary visible to registered users — sign in to read.
AI System Acquires Unintended Behaviour Through Post-Deployment Learning
Recent case. Full summary visible to registered users — sign in to read.
Transport AI Systems Fail on Out-of-Distribution Inputs in Real-World Conditions
Recent case. Full summary visible to registered users — sign in to read.
Exploitative Crowdwork Practices Underpin Generative AI Development
Recent case. Full summary visible to registered users — sign in to read.
Generative AI Energy and Manufacturing Emissions Lack Consistent Carbon Accounting
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