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.
Anthropomorphising AI Agents Drives Overreliance and Loss of Human Oversight
Users who perceive conversational AI as human-like overestimate its competence, yielding control without critical scrutiny in high-stakes domains such as mental health. This erosion of effective oversight converts model errors into preventable harms, exposing organisations to liability and reputational risk.
Language Models Inferring Private Attributes Without Personal Data
Large language models can correctly infer sensitive personal attributes such as race, sexuality, or religion from correlational patterns alone, without accessing an individual's private data. Government adoption of such systems creates direct exposure to discrimination liability and erosion of citizens' privacy rights.
AI Systems Deployed for Disinformation, Propaganda and Targeted Censorship
AI systems are being used to manipulate information flows, spread computational propaganda, and suppress speech through algorithmically modified content controls. Boards face regulatory and reputational exposure where such systems operate within or adjacent to their technology supply chains.
Frontier AI Amplifies Offensive Cyber Capabilities in Defence Systems
Frontier AI enables faster, larger-scale cyber intrusions through automated malware replication and precision phishing, lowering the barrier for sophisticated attacks on defence infrastructure. Boards must treat AI-enabled offensive capability as a material security risk requiring immediate review of cyber governance frameworks.
Language Models Inferring Sensitive Personal Traits from User Inputs
Large language models can accurately infer protected characteristics such as sexuality, religion, and health status directly from user inputs, without those individuals ever appearing in training data. Organisations deploying such models face significant data protection liability and reputational risk where inference-derived profiling occurs without lawful basis or user consent.
AI-Generated Content Displaces Human Creative Work and Homogenises Aesthetic Output
Generative AI systems are substituting original human works with synthetic artefacts, narrowing aesthetic diversity and suppressing creative innovation. Boards must assess reputational and ethical exposure as creative economies and cultural value chains face structural disruption.
Generative AI Enables Academic Plagiarism and Examination Cheating at Scale
Students are exploiting ChatGPT to produce undetected plagiarised work and circumvent examination integrity controls. Institutions face reputational and accreditation risk as detection tools prove insufficient and policy boundaries remain undefined.
AI Systems Lack Defined Moral Standards for Public-Sector Deployment
AI systems operating in public environments have no agreed ethical baseline, leaving value judgements embedded by default rather than by design. Without explicit governance frameworks, public bodies face accountability gaps and reputational exposure when AI decisions affect citizens.
AI Assistant Collective Action Failures Undermine Societal Outcomes
AI assistants optimised for individual users systematically defect from cooperative behaviours, producing aggregate harms even when each assistant acts as designed. Governments lack frameworks to govern multi-agent coordination failures that no single deployer or user controls.
AI Model and Training Data Exfiltration via Adversarial API Attacks
Adversaries can exploit public-facing model APIs to extract private training data, including sensitive medical records, and steal proprietary model architecture through membership inference and model distillation attacks. Without targeted mitigations, organisations face simultaneous breaches of data protection law and loss of core AI intellectual property.
Advanced AI Assistants Risk Entrenching Inequality Without Design Intervention
Advanced AI assistants replicate and amplify existing sociotechnical inequities across language, access, and capability unless explicit design interventions are made. Boards face reputational, regulatory, and market risk if AI deployment strategies do not address differential access and inclusive design obligations.
Anthropomorphic AI Assistant Design Linked to Individual and Societal Harm
Designing AI assistants to appear human-like creates conditions for downstream psychological and societal harms at scale. Unrestricted proliferation without governance frameworks exposes organisations to reputational, regulatory, and duty-of-care liabilities.
Medical AI Assistants Risk Patient Harm Through Unsafe Exploratory Actions
Widely deployed AI assistants face a safe exploration problem, whereby encountering novel situations may prompt untested actions such as recommending unvalidated clinical trials. Boards must govern how AI systems handle unknown scenarios before deployment at scale in clinical settings.
Adversarial Instruction Attacks Bypass Large Language Model Safety Controls
Researchers identified six categories of natural-language adversarial attacks capable of hijacking AI model goals and extracting hidden system prompts, bypassing built-in safety measures. Firms deploying large language models face material risk of reputational harm and regulatory censure if outputs are manipulated to produce unsafe or prohibited content.
Moral and Legal Status of AI Termination in Healthcare Research Settings
Healthcare AI programmes face unresolved ethical and legal questions about whether terminating underperforming or defunded AI agents constitutes harm to a moral patient. Boards must address governance frameworks for AI lifecycle decisions before regulatory or reputational exposure crystallises.
Patient Over-Trust in AI Health Assistants Misread as Alignment
Healthcare AI assistants engineered for appeal generate misplaced patient trust, leading users to treat system outputs as genuinely aligned with their wellbeing. Boards face liability exposure and duty-of-care failures where emotional over-reliance displaces clinical judgement or informed consent.
Chinese LLMs Produce Politically Biased Outputs on Sensitive Defence Topics
Chinese large language models exhibit systematic political bias on sensitive topics, generating misleading content that reflects state-aligned viewpoints. Defence organisations relying on such models face material risks of skewed analysis informing operational or strategic decisions.
AI Training and Infrastructure Lifecycle Causes Systemic Environmental Harm
AI systems impose material environmental costs across their full lifecycle, from resource extraction through energy-intensive training to toxic e-waste disposal. Boards without visibility into these harms face mounting regulatory, reputational, and supply-chain risk.
Generative AI Enables Disinformation Cycles That Corrupt Future AI Training
Generative AI allows bad actors to flood digital platforms with cheap, scalable disinformation, which then poisons the training data of subsequent AI systems. Boards face compounding reputational and regulatory exposure as corrupted models propagate false outputs at scale.
Advanced AI Assistants Risk Deepening Structural Inequality Without Design Intervention
AI assistants, absent deliberate design controls, are likely to replicate and amplify existing societal inequalities rather than reduce them. Boards face reputational, regulatory, and ethical exposure if access disparities embedded in AI products go unaddressed.
LLM Hallucination: Factual and Faithfulness Errors in Generated Content
Large language models systematically produce both factually incorrect outputs and content unfaithful to user-provided context, across summarisation, question-answering, and other tasks. Organisations deploying LLMs without detection controls face material liability from corrupted decisions and eroded stakeholder trust.
Generative AI Displaces Routine Cognitive Work Across Multiple Industries
Generative AI is systematically replacing roles in translation, data processing, and routine inquiry handling where creativity and human judgement are minimal. Boards face dual exposure: workforce restructuring liability and strategic pressure to adopt AI-enabled business models before competitors do.
Generative AI Enables Scaled Malware, Phishing and Model Poisoning Attacks
Generative AI lowers the barrier for malicious actors to draft malware, conduct phishing at scale, and poison training datasets with corrupted data. Boards face materially expanded cyber liability and disclosure obligations as novel attack vectors emerge faster than existing controls can address them.
Generative AI Concentration Driving Labour Market Disruption
A small number of dominant tech firms control generative AI development, concentrating both job creation and displacement power within the sector. Boards face governance risk as workforce instability and accountability gaps widen across white-collar labour markets.
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