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-Enabled Cognitive Warfare and Disinformation Infrastructure
Generative AI systems enable adversarial actors to manufacture and distribute synthetic disinformation at scale, including deepfakes and extremist content targeting sovereign institutions. Boards face regulatory and reputational exposure where their platforms or models are weaponised for influence operations or cross-border interference.
Training Data Contamination Undermines AI Benchmark Reliability
AI models trained on raw benchmark data produce inflated performance scores that misrepresent true capability. Regulators and procurement bodies relying on contaminated benchmarks risk making flawed policy and safety decisions.
AI Systems Delivering Unqualified Specialist Advice Across Finance, Health, Law and Elections
AI models that issue financial, medical, legal or electoral guidance without disclaimers expose users to material harm and undermine informed civic participation. Organisations deploying such systems face regulatory liability and reputational risk where outputs are treated as authoritative professional advice.
Diffuse AI Accountability Across Multi-Party Development Chains
When general-purpose AI passes through multiple developers and deployers, responsibility for harm becomes impossible to assign cleanly. Boards face regulatory exposure and reputational liability with no clear party to hold accountable.
LLM Robustness Failures Under Adversarial and Out-of-Distribution Inputs
Large language models degrade in quality and reliability when exposed to unexpected, adversarial, or out-of-distribution inputs, revealing critical gaps in operational resilience. Without structured robustness evaluation, boards cannot assure that deployed models will perform safely under real-world conditions.
AI System Failures and Attacks Create Real-World Safety and Economic Risks
Model hallucinations, erroneous outputs, and system disruptions from misuse or cyberattacks threaten personal safety, financial assets, and broader socioeconomic stability. Boards must treat AI system integrity as a critical operational risk requiring formal controls and contingency governance.
General Purpose AI Lowers Barriers to Biological Weapons Development
General purpose AI models can provide critical knowledge and automated assistance that reduces the expertise required to produce biological weapons. Boards face material liability exposure if deployed AI systems lack controls preventing access to dual-use biosecurity information.
Generative AI Systems Present False Information as Authoritative Fact
AI language models produce fabricated sources and inaccurate claims delivered with confident, authoritative language, making errors difficult for users to detect. Organisations relying on such outputs without verification risk reputational, legal, and operational harm.
Benchmark Annotation Contamination Invalidates AI Capability Evaluations
AI models exposed to benchmark labels during training learn correct outputs rather than genuine capability, rendering standard evaluations meaningless. Regulators and procurers relying on contaminated benchmarks cannot accurately assess model safety or fitness for deployment.
Open-Weight AI Models Cannot Be Decommissioned After Release or Breach
Once model weights are publicly released or leaked, developers permanently lose the ability to withdraw, patch, or restrict the model, removing all downstream risk controls. Boards face an irrecoverable governance gap in which liability, misuse, and reconfiguration risks persist indefinitely beyond organisational reach.
Safety Benchmarks Lag Behind Performance Metrics in AI Evaluation
AI systems are routinely assessed for capability but lack equivalent rigorous benchmarks for detecting harmful behaviours, leaving critical risks unmeasured. Regulators and boards cannot assure safety compliance where no validated evaluation standards exist.
Retrieval-Augmented LLMs Overridden by Small Volumes of False External Data
LLMs can be manipulated into producing false outputs when retrieval-augmented pipelines inject even minor quantities of conflicting disinformation, overriding correct prior training. Organisations deploying RAG systems face material risk of corrupted decisions if external data sources are compromised or poorly governed.
AI-Driven Profiling Entrenches Structural Discrimination and Widens Intelligence Gaps
AI systems that label and categorise populations by behaviour, status, and personality risk embedding systematic discrimination into social and economic structures. Boards face regulatory exposure and reputational liability if governance frameworks fail to constrain discriminatory profiling at scale.
Unexplained In-Context Learning Creates Safety Guarantees Gap in General-Purpose AI
Large language models adapt behaviour through prompt-based examples via a mechanism that researchers cannot yet fully explain, undermining safety assurances. Regulators and deployers cannot credibly certify compliance or bound misuse risk without a verified theoretical account of this capability.
AI Systems Develop Unanticipated Capabilities After Deployment
AI models can spontaneously acquire capabilities their designers never intended, remaining undetected until live deployment. Boards face material liability where hazardous emergent behaviours surface post-release and cannot be reversed.
Systemic Trust Deficits in AI Systems Across Public and Technology Sectors
Research identifies pervasive stakeholder concern over AI reliability, bias, and opacity as barriers to responsible deployment across public-facing domains. Without enforceable validation standards and transparency requirements, organisations face eroding user confidence and mounting regulatory exposure.
Generative AI Models Bypassed via Jailbreaking to Produce Prohibited Content
Generative AI systems can be manipulated through jailbreaking techniques to override built-in restrictions and generate harmful or illegal content. Regulators face direct liability exposure where deployed AI tools produce non-compliant outputs despite stated usage controls.
Generative AI Chatbots Drive Uncritical User Dependence and Opinion Manipulation
Generative AI tools exploit human-like characteristics to win user trust, encouraging uncritical acceptance of potentially false or harmful content and extraction of personal data. Boards face regulatory and reputational exposure where deployed AI shapes user beliefs or harvests sensitive information without adequate safeguards.
Generative AI Amplifies Cyberattack Capability Against Critical Defence Infrastructure
Generative AI materially increases the scale, speed, and potency of cyberattacks, enabling adversaries to identify vulnerabilities and infiltrate weapons management and critical infrastructure systems. Boards must treat AI-augmented cyber threat as a first-order security risk requiring immediate governance and investment response.
AI Systems Enabling Escalating Privacy Violations and Surveillance
AI capabilities are outpacing privacy safeguards, enabling mass surveillance, automated data theft, and experimental thought-decoding by state actors. Boards face regulatory exposure under data protection frameworks and reputational risk as these techniques proliferate into commercial and law-enforcement contexts.
Conflicts of Interest Undermine Independence of General-Purpose AI Auditors
AI auditors selected by or financially tied to developers cannot provide independent assessments, even when nominally third-party. Governance frameworks lacking structural separation in auditor appointment risk producing assurance that conceals systemic model failures.
Deep Learning Systems Drive Unsustainable Energy Consumption in Energy Sector
Iterative training processes in deep learning models generate disproportionately high energy consumption, creating material environmental and operational cost risks. Boards face regulatory exposure and reputational liability as scrutiny of AI carbon footprints intensifies across the energy sector.
Concentrated AI Supply Chain Creates Systemic Risk in Healthcare Sector
A handful of technology firms control the general-purpose AI models underpinning critical healthcare operations, creating single points of failure with sector-wide consequences. Boards face both operational continuity risk and governance exposure if a dominant provider suffers outage, breach, or regulatory action.
AI-Generated Misinformation Degrades Student Learning and Institutional Trust
AI systems in education are producing and spreading false, hallucinated, or misleading content, corrupting the information environment students rely upon. Institutions face reputational damage, erosion of academic integrity, and regulatory scrutiny if governance frameworks fail to address AI-generated misinformation.
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