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.
ChatGPT and Clearview AI Face GDPR Bans and Mass Data Collection Scrutiny
Regulators banned ChatGPT in Italy over a data breach and unlawful training-data collection, while Clearview AI scraped three billion images without consent. Boards face material compliance exposure as GDPR enforcement spreads across France, Ireland, and Germany.
Generative AI Tools Expose Firms to IP Liability Through Unlicensed Training Data and Code
AI models trained on copyrighted content and open-source code are generating outputs that breach intellectual property rights, including viral licences such as GPL. Firms adopting AI coding assistants risk injunctions, compelled disclosure of proprietary source code, and litigation from rights holders.
Generative AI Hallucination Produces Misleading Outputs
Large language models systematically generate plausible but false or nonsensical content, presenting it with unwarranted confidence. Organisations deploying these systems without robust output verification face material liability and reputational harm from decisions made on fabricated information.
Pre-Trained Language Models Memorise and Expose Personal Data
Large language models trained on internet corpora retain and can reproduce personal data including phone numbers, email addresses, and home addresses. Organisations deploying such models face regulatory liability and reputational harm if memorised personal data is surfaced through user queries.
Biased Training Data Propagates Discrimination Through UN AI Systems
AI systems trained on historically biased data will reproduce and scale those biases in outputs and decisions. Organisations deploying AI without rigorous data audits face reputational, legal, and ethical failures at institutional scale.
Demeaning social groups — case from Sociotechnical Harms of Algorithmic Systems: Scoping a Taxonomy for Harm Reduction
Demeaning of social groups to occur when they are when they are “cast as being lower status and less deserving of respect"... discourses, images, and language used to marginalize or oppress a social group...
AI Hardware Demand Accelerates Critical Mineral Depletion Risk
Rapid scaling of AI infrastructure is consuming nickel, cobalt, and lithium at rates that may outpace global reserves within foreseeable timescales. Boards face material supply chain exposure and reputational liability if hardware procurement strategies do not account for resource sustainability constraints.
Algorithmic erasure of minority social groups through biased training data
Systematic under-representation of minority social groups in training data causes AI systems to render certain identities and experiences invisible in outputs. Organisations deploying such systems face reputational, legal, and equality-duty exposure as discriminatory design choices become embedded at scale.
Algorithmic Stereotyping of Social Groups in Language Model Outputs
Language models embed and reinforce group stereotypes through subtle output patterns, such as treating professional roles as implicitly gendered. Organisations deploying these systems face reputational, legal, and regulatory exposure if outputs reflect discriminatory assumptions at scale.
AI System Discrimination Against Protected Groups at the United Nations
Poorly designed AI systems at the UN risk embedding discriminatory bias against protected groups into institutional decision-making. Boards must ensure ethical design standards and bias audits are mandated before any AI deployment.
Generative AI Weaponisation and Existential Safety Risks in Defence Contexts
Generative AI models present compounding defence risks, from autonomous shutdown evasion and unforeseen capability emergence to active misuse by hostile actors for biological weapons ideation. Boards must mandate red-teaming protocols, restrict open-source model access, and prioritise safety governance before capability development accelerates further.
Image Tagging System Erases Social Group Identity from Tagged Content
An algorithmic image tagging system failed to recognise socially significant group identities, rendering affected users invisible within their own content. This structural exclusion creates reputational and equity risk for organisations deploying such systems at scale.
Algorithmic Systems Withholding Healthcare Resources from Marginalised Groups
AI allocation tools systematically deny healthcare information and resources to historically marginalised populations, compounding existing inequalities in material outcomes. Boards face regulatory exposure and reputational liability where procurement decisions embed discriminatory access patterns into clinical workflows.
Automated Gender Classifiers Force Binary Categories on Non-Binary Users
Algorithmic classification systems impose binary gender labels on individuals who identify outside that framework, removing their autonomy over self-disclosure. Organisations deploying such systems face material equality, regulatory, and reputational exposure under UK equalities law.
Fragmented AI Governance Leaves Frontier Model Risks Unsanctioned
Generative AI outpaces legal frameworks, with no binding international safety standards or effective mechanisms to sanction non-compliant frontier model developers. Governments without regulatory access to AI lab processes cannot assess systemic risk, exposing public institutions to ungoverned technological harm.
Generative AI Systems Weaponised for Non-Consensual Content at Scale
Diffusion models and advanced AI assistants enable automated creation of non-consensual imagery, deepfakes, and targeted harassment at unprecedented scale. Technology firms face material regulatory exposure and reputational liability as legislators accelerate intervention against platform-enabled AI misuse.
Generative AI Labour Displacement and Workforce Deskilling Risk
Generative AI is displacing workers across customer service, software engineering, and crowdwork platforms whilst deskilling remaining staff through automation of core tasks. Boards face reputational, regulatory, and productivity risks as socioeconomic inequality widens and workforce capability erodes.
AI-Enabled Personal Data Profiling for Targeted Advertising
AI systems create systematic incentives to exploit personal data for behavioural profiling and precision advertising without meaningful consent. Organisations that fail to govern these uses face regulatory exposure and lasting reputational damage with customers and partners.
LLM Use in Academic Writing Erodes Integrity and Pollutes Scientific Literature
Generative AI homogenises student writing, suppresses individual expression, and floods scientific publishing with low-quality synthetic manuscripts. Institutions face compounding risks to academic standards, credential credibility, and research corpus integrity.
AI-Enabled Malware and Hacking Identified as Operational Threat
AI is actively lowering the barrier for cybercriminals, enabling more sophisticated malware and hacking operations at scale. Boards must treat AI-augmented cyber threats as a material risk requiring urgent review of security controls and incident response frameworks.
Lethal Autonomous Weapons Raise Accountability Gap in Defence AI Governance
AI-driven weapons capable of autonomous lethal targeting operate without clear legal or ethical accountability frameworks at international level. Defence organisations face material governance exposure as regulatory consensus remains unresolved at the United Nations.
Deepfake Technology Enables Large-Scale Deceptive Content Creation
AI systems now generate synthetic text, images, audio, and video indistinguishable from authentic content by human observers. Boards face material reputational, legal, and market-integrity risks as verification frameworks struggle to keep pace with generative capabilities.
AI Decision Delegation Strips Human Autonomy in High-Stakes Contexts
Opaque, uncontestable AI systems remove meaningful human agency when consequential decisions are delegated to machines. Boards face reputational and legal exposure where affected individuals cannot challenge or understand automated outcomes.
Unintended AI Failure Modes and Developer Liability
Advanced AI systems can produce harmful outcomes through unintended failure modes attributable to the system or its developer. Boards face direct liability exposure where governance frameworks fail to establish clear accountability for AI-induced accidents.
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