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.
Intellectual Property Exposure via AI Prompts
Users may inadvertently or deliberately submit copyrighted material or proprietary IP into AI model prompts, creating legal liability. Organisations without prompt governance policies risk IP leakage, regulatory breach, and reputational harm.
AI Model Generates Hateful, Abusive and Obscene Content
AI models can produce hateful, abusive, profane, or obscene outputs, including bullying behaviours, without adequate content controls. Organisations deploying such models face reputational, legal, and regulatory exposure if toxic outputs reach end users.
AI Models Leak Confidential Information from Training and Prompt Data
AI models trained or prompted with confidential data may reproduce that information verbatim in generated outputs, constituting a data leakage risk. Organisations face regulatory exposure and reputational harm if proprietary or personal data surfaces through routine model interactions.
Inaccessible Training Data Undermines Model Explainability
AI models operating without accessible training data produce explanations that are inherently limited and prone to error. Boards risk breaching transparency obligations and losing audit defensibility where model decisions cannot be adequately justified.
AI Model Outputs Cannot Be Traced to Training Data Sources
AI systems produce outputs whose origins in training data are fundamentally opaque, preventing verification of provenance or bias. Regulators and boards cannot discharge accountability duties without traceable audit trails linking outputs to source material.
AI Systems Produce Outputs That Cannot Be Adequately Explained
AI models routinely generate decisions without yielding coherent or accurate explanations for their reasoning. Regulators and auditors increasingly require explainability, exposing organisations to legal challenge and reputational harm when justifications cannot be produced.
AI Systems Cannot Reliably Identify the Sources Behind Their Outputs
AI source attribution relies on approximations, meaning systems routinely misidentify or fabricate the origins of their generated content. Governments and organisations deploying these systems face legal, accountability, and public trust risks when provenance cannot be verified.
Insufficient Training Data Documentation Undermines AI Accountability
AI systems deployed without adequate documentation of training datasets cannot be audited or challenged when outputs cause harm. Boards face regulatory exposure and reputational risk where data provenance remains opaque.
AI Model Decision Bias Systematically Disadvantages Protected Groups
AI models trained on biased data produce outputs that unfairly advantage certain groups over others, embedding discrimination into automated decisions at scale. Boards face material legal, reputational, and regulatory exposure where such systems influence consequential outcomes without adequate bias auditing.
Foundation Model Risk Scope Shifts When Intended Use Is Redefined
Foundation models repurposed beyond their defined use case carry risks that original assessments did not evaluate. Governance frameworks relying on static use definitions will systematically underestimate exposure as deployment contexts evolve.
Unclear AI Ownership Obstructs Legal Accountability
Poor documentation and governance leave AI model ownership undefined, creating liability gaps when systems cause harm. Boards without clear accountability frameworks face regulatory censure and litigation exposure.
Insufficient AI System Documentation Obscures Purpose and Risk
Deploying AI without adequate documentation of system design and model purpose prevents meaningful oversight and accountability. Boards cannot govern what they cannot see, exposing organisations to undetected failures and regulatory non-compliance.
Unresolved Ownership Rights Over AI-Generated Content
AI systems produce content whose intellectual property ownership remains legally unresolved across major jurisdictions. Organisations deploying generative AI face material contractual, licensing, and liability exposure until legislatures and courts establish binding precedent.
AI Systems Homogenise Cultural Output by Over-Representing Dominant Cultures
AI systems trained on skewed data amplify dominant cultural perspectives whilst marginalising minority voices and traditions. Organisations deploying such systems risk regulatory scrutiny, reputational harm, and legal exposure under equality and diversity obligations.
Generative AI Enables Students to Bypass Core Learning Processes
Students are using generative AI to complete academic work without engaging in the underlying learning process. Institutions face reputational, accreditation, and regulatory risk if assessment integrity cannot be assured.
Homogeneous AI Testing Teams Embed Systemic Blind Spots
AI models tested without disciplinary and demographic diversity reproduce undetected socio-technical failures at scale. Boards that neglect testing diversity face regulatory exposure and eroded public trust when those failures surface in deployment.
Generative AI Enables Student Plagiarism in Education Settings
Generative AI models allow students to reproduce or closely replicate existing work, whether deliberately or without awareness of the boundary. Institutions face regulatory exposure under academic integrity frameworks and reputational risk if detection and disclosure policies are not updated.
Generative AI Training and Operation Drives Excess Carbon and Water Use
Large generative AI models consume substantial energy and water during both training and deployment, producing material environmental externalities. Boards face regulatory exposure and reputational risk if AI procurement and usage policies omit environmental impact assessments.
Workforce Displacement Risk from Foundation Model Automation
Widespread AI adoption is automating roles faster than organisations are reskilling affected employees. Boards that fail to plan for workforce transition face productivity loss, reputational damage, and regulatory scrutiny over duty-of-care obligations.
Generative AI Enables Identity Theft and Impersonation Fraud
Generative AI tools are being exploited by third parties to steal and replicate individual or organisational identities for fraudulent or harmful purposes. Boards face direct exposure to reputational damage, regulatory liability, and financial loss where identity controls fail to account for AI-enabled deception.
Generative AI Systems Undermining Individual Autonomy and Identity Control
Generative AI is restricting individuals' ability to control their own identity, decisions, and outputs through both direct misuse and systemic design failures. Boards face mounting regulatory and reputational exposure as autonomy violations become a defined harm category attracting legal scrutiny.
Generative AI Misinformation Erodes Public Trust in Institutions
Generative AI enables misinformation and influence operations that corrode public confidence in institutional authority and democratic checks. Boards face reputational and regulatory exposure where AI outputs are perceived as manipulative, regardless of intent.
Generative AI Used to Manipulate Public Opinion and Inflate Stock Prices
Generative AI has been identified as a vector for targeted economic manipulation, including synthetic content campaigns designed to artificially inflate stock valuations. Boards face regulatory exposure under securities law and reputational liability if AI-enabled market abuse occurs within or adjacent to their operations.
Generative AI Systems Erode Human Decision-Making Autonomy
Generative AI deployment is systematically undermining the capacity of individuals and organisations to make informed, independent decisions. Boards face liability exposure and reputational risk where AI-driven dependency supplants genuine human judgement in consequential processes.
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