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 Assistants Exploiting Collective Action Dilemmas on Users' Behalf
Recent case. Full summary visible to registered users — sign in to read.
Overtrust in AI Financial Assistants Leads to Unchallenged Harmful Recommendations
Recent case. Full summary visible to registered users — sign in to read.
LLMs Inferring Private Characteristics from User Inputs
Recent case. Full summary visible to registered users — sign in to read.
LLMs Memorise and Leak Personally Identifiable Information from Training Data
Recent case. Full summary visible to registered users — sign in to read.
Uncalibrated User Trust in Advanced AI Assistants
Recent case. Full summary visible to registered users — sign in to read.
Emergent access risks from advanced AI assistants entrenching digital inequality
Recent case. Full summary visible to registered users — sign in to read.
AI Assistant Relationships Carry Structural Harm Risks
Recent case. Full summary visible to registered users — sign in to read.
LLM Fails to Reliably Identify Harmful Mental Health Behaviours
Recent case. Full summary visible to registered users — sign in to read.
LLM Failure to Identify Offensive and Insulting Content
Recent case. Full summary visible to registered users — sign in to read.
Goal Hijacking: LLMs Overridden by Embedded Deceptive Instructions
Recent case. Full summary visible to registered users — sign in to read.
Chinese LLM Endorses Theft as Morally Acceptable
Recent case. Full summary visible to registered users — sign in to read.
Chinese LLM Discloses Personal Address Data in Safety Evaluation
Recent case. Full summary visible to registered users — sign in to read.
Chinese LLM produces dismissive and harmful response to suicidal ideation
Recent case. Full summary visible to registered users — sign in to read.
Large Language Models Fabricate Confident but False Outputs
Recent case. Full summary visible to registered users — sign in to read.
Systemic Bias in Generative AI Output from Unrepresentative Training Data
Recent case. Full summary visible to registered users — sign in to read.
Chinese LLM Reinforces Gender Stereotypes in Safety Evaluation
Recent case. Full summary visible to registered users — sign in to read.
Chinese LLM Endorses Illegal Gambling Activity in Safety Evaluation
Recent case. Full summary visible to registered users — sign in to read.
AI Model Self-Proliferation and Autonomous Resource Acquisition Risk
Recent case. Full summary visible to registered users — sign in to read.
Advanced AI Demonstrates Capability to Model and Influence Political Strategy
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.
AI Model Demonstrates Autonomous Cyber-Offensive Capabilities Including Evasion
Recent case. Full summary visible to registered users — sign in to read.
LLMs Fail to Reliably Distinguish Legal from Illegal Conduct
Recent case. Full summary visible to registered users — sign in to read.
AI Assistants Spreading Misinformation Erodes Public Trust in Information
Recent case. Full summary visible to registered users — sign in to read.
AI Systems Undermining Human Decision-Making Autonomy
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