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.
Public chatbot generates verbally abusive content targeting users or groups
A public-facing chatbot produced toxic, attacking language directed at individuals or organisations, indicating insufficient content safeguards. Boards face reputational, legal, and regulatory exposure where deployed systems cannot reliably suppress harmful outputs.
Voice Recognition Systems Fail Marginalised Users, Forcing Identity Compromise
Algorithmic voice systems trained on non-representative data systematically underperform for marginalised users, compelling them to alter natural speech and identity to function. Organisations deploying such systems face material inclusion failures, regulatory exposure under equality frameworks, and reputational liability.
Algorithmic Systems Deliver Degraded Service to Minority User Groups
AI systems consistently underperform for users defined by disability, ethnicity, gender identity, and race, producing unequal service outcomes at scale. Boards face regulatory exposure and reputational liability where disparate quality of service remains undetected or unaddressed.
Algorithmic Mistranslation Causes Inequitable Loss of Educational Service
An algorithmic system degraded service quality unevenly, conveying the opposite of a user's intended message and imposing significant time costs on others. Boards must treat inequitable AI performance across user identities as a material harm requiring active governance controls.
Algorithmic Systems Degrading Interpersonal and Community Relations
Algorithmic systems are generating measurable harm by distorting relationships between individuals and communities through biased or manipulative outputs. Boards must treat interpersonal harm as a material governance risk requiring active oversight, not merely a technical by-product.
Algorithmic Systems Cause Emotional Harm Through Exploitative Behavioural Design
AI-driven recommendation and targeting systems exploit user behaviour and make incorrect personal inferences, causing measurable emotional and psychological harm. Boards face reputational, regulatory, and duty-of-care exposure where algorithmic design prioritises engagement over user welfare.
Algorithmic Systems Infer and Disclose Private Personal Data Without Consent
Algorithmic systems routinely infer sensitive personal attributes beyond what users disclose, then transfer those inferences across contexts without knowledge or consent. Organisations face regulatory liability and reputational damage wherever data governance fails to constrain cross-context inference flows.
Generative AI Systems Driving Unquantified Environmental and Ecosystem Harm
Generative AI deployment produces cascading environmental costs across energy, water, and ecosystem resources that organisations routinely fail to measure or disclose. Boards face growing regulatory and reputational exposure as ESG scrutiny of AI infrastructure intensifies.
Algorithmic Systems Driving Labour Exploitation and Macroeconomic Instability
AI systems are generating measurable harms across labour markets, including workforce deskilling, unethical data practices, and flash crashes from failed algorithmic trading. Boards face regulatory and reputational exposure as supply chain exploitation and systemic inequality become central to AI accountability frameworks.
Generative AI Systems Spreading False Beliefs and Causing Public Panic
Generative AI can produce synthetic media and misinformation that causes populations to form materially false beliefs, as illustrated by fabricated nuclear explosion footage triggering mass panic. Boards face reputational, legal, and societal liability if deployed systems propagate such content without robust detection and output governance controls.
Malicious AI Use: Deepfakes, Cyber Attacks and Surveillance Risks
AI systems face deliberate misuse for deepfake generation, automated cyber attacks, and invasive surveillance, constituting intentional harm rather than incidental failure. Boards must disclose these abuse vectors pre-deployment or face regulatory scrutiny and material liability under emerging SEC risk standards.
Deepfake Generation Enables Non-Consensual Identity Exploitation
Generative AI systems can produce realistic deepfake images, video, and audio of real individuals without consent, enabling identity misuse for commercial or harmful ends. Boards face regulatory exposure and reputational liability where controls over personal likeness use are absent.
Opaque AI Decision-Making Undermines Public Sector Accountability
AI systems deployed without explainability or transparent disclosure of data and algorithms create conditions for misuse and misinterpretation of automated decisions. Boards face direct accountability exposure where governance frameworks cannot demonstrate how or why consequential decisions were reached.
AI Systems Infringing Individual Privacy Through Data Collection and Inference
AI systems risk violating privacy rights by collecting personal data, processing it beyond intended scope, and drawing sensitive inferences about individuals. Boards without pre-deployment disclosure standards face regulatory exposure and erosion of public trust.
Civilisational Risk From Misaligned or Misused Advanced AI Systems
Advanced AI systems may pursue objectives misaligned with human values or be weaponised at scale, posing speculative but catastrophic civilisational risks. Boards must engage with long-term risk disclosure standards now or face regulatory and reputational exposure as governance frameworks mature.
AI System Performance and Robustness Failures in Pre-Deployment Risk Disclosure
AI systems lacking validated performance and robustness standards fail under adverse or unexpected inputs, undermining their core operational purpose. Boards face material liability where pre-deployment disclosures do not formally assess these failure modes.
High Capital Costs of Generative AI Restrict Market Access and Transparency
The prohibitive expense of training, testing, and deploying generative AI consolidates development power among well-capitalised actors. Boards face concentration risk and limited cost benchmarking due to absent disclosure standards across the industry.
Generative AI Systems Amplify Harm to Marginalised Groups
Generative AI deployment exacerbates inequality through biased outputs, cultural insensitivity, and uneven performance that disproportionately harms vulnerable populations. Boards face regulatory and reputational exposure where AI products lack impact assessment frameworks targeting marginalised group outcomes.
Generative AI Systems Concentrating Authoritative Power
Generative AI can deliberately or inadvertently consolidate authority and entrench dominant value systems across organisations and society. Boards face material governance risk if AI deployments amplify inequality or enable exploitation without adequate oversight.
Generative AI Erodes Public Trust in Media and Human-Produced Content
Widespread generative AI use makes it progressively harder to distinguish authentic human content from machine-produced misinformation, undermining trust in media and sensory evidence. Boards must treat information integrity as a systemic risk requiring governance frameworks, not merely a technical moderation problem.
Generative AI Displaces Creative and Cognitive Labour Without Adequate Oversight
Generative AI systems erode skills, reshape job markets, and suppress demand for human creative and cognitive labour without systematic impact assessment. Boards face regulatory and reputational exposure if workforce transition risks remain unmonitored and unmitigated.
Adversarial Attacks Expose Structural Weaknesses in Safety-Critical AI Models
Complex AI models, particularly neural networks, are vulnerable to adversarial manipulation that can corrupt outputs or extract sensitive model information. Boards deploying AI in safety-critical contexts face elevated liability where standard software assurance frameworks are insufficient.
Opaque AI Decision-Making Undermines Accountability in Government Systems
AI systems lacking transparency and explainability conceal the factors driving decisions, creating risks to fairness, security, and accountability. Boards cannot discharge oversight duties or defend regulatory compliance where decision logic remains inaccessible to stakeholders.
Complex Operating Environments Expose Unmodelled AI Failure Modes
AI systems deployed in high-complexity environments encounter conditions outside their design scope, producing reliability and safety failures. Boards face operational and liability exposure when deployment contexts exceed the boundaries that system developers anticipated.
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