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.
Adversarial Training Produces Models That Become Less Robust Over Time
Models hardened against adversarial attacks can deteriorate in resilience as training progresses, leaving deployed systems more vulnerable than testing indicated. Organisations relying on adversarial training as a security assurance measure may hold false confidence in their AI defences.
Accountability Gaps in AI Decision-Making Systems
AI systems cannot reliably replicate the contextual, moral, and empathetic dimensions of human accountability, leaving automated decisions structurally ungovernable. Governments deploying AI in public services face legal exposure and democratic legitimacy risks where no accountable agent can be identified.
Autonomous Weapons Systems Outperform Human Pilots, Raising Lethal AI Governance Gaps
AI agents now exceed experienced combat pilots in simulated aerial engagements, and fully autonomous lethal drones are already operational without mandatory human oversight frameworks. Boards with defence exposure face material regulatory and liability risk as international governance for autonomous weapons remains absent.
AI Development Triggers Resource Conflicts Over Data Centres and Semiconductors
The rapid scaling of AI infrastructure creates geopolitical and physical conflict risks centred on data centres, semiconductor facilities, and critical raw materials. Boards must treat AI supply chain concentration as a material strategic and operational risk requiring active oversight.
Black-Box AI in Healthcare Blocks Clinical Verification
AI systems used in medical decision-making cannot be verified due to their opaque, non-linear structures, leaving clinical outputs without audit trails. Regulators and boards face direct liability where unverifiable AI informs patient care.
Frontier AI Systems Amplify Discriminatory Bias Across All Output Modalities
Frontier AI models trained on internet-scale data systematically reproduce and amplify misogynistic, ageist, and white supremacist content across text, image, and other generative outputs. Organisations deploying these systems face material reputational, legal, and regulatory exposure if bias assurance is absent from procurement and governance frameworks.
AI Systems in Elder and Child Care Risk Psychological Manipulation
Advanced AI deployed in elder and child care settings presents documented risks of psychological manipulation and clinical misjudgement of vulnerable users. Boards face mounting liability exposure and regulatory scrutiny where duty-of-care obligations intersect with autonomous system deployment.
AI Systems Exposed to the Same Cyber Vulnerabilities as Conventional Software
Learning systems carry identical attack surfaces to standard software, making them susceptible to exploitation, data poisoning, and adversarial manipulation. Boards must treat AI components as critical cyber assets within existing security governance frameworks.
Autonomous Vehicle AI Creates Unresolved Liability and Ethical Decision Gaps
Autonomous transport AI lacks settled frameworks for allocating accident liability and encoding ethical decision logic in life-critical scenarios. Governments and operators face regulatory exposure and public trust risk until clear accountability structures are legislated.
Generative AI Enables Scalable Mass Surveillance of Individuals
Generative AI drastically reduces the cost and complexity of monitoring behaviour, beliefs, and communications at population scale in real time. Boards must assess exposure to regulatory, reputational, and human rights liability where their technology is deployed in surveillance contexts.
Absence of Accountability Frameworks in AI Decision-Making
AI systems making consequential decisions without clear procedural or substantive standards create governance vacuums where responsibility cannot be assigned. Boards face direct liability exposure when no accountable party can be identified following a harmful or non-compliant AI output.
Generative AI Displacing Skilled Workers and Concentrating Economic Power
AI systems designed to replicate human capabilities risk displacing expert workers, suppressing wages, and concentrating wealth among capital owners. Boards face regulatory and reputational exposure as workforce inequality intensifies and governance frameworks struggle to keep pace.
Autonomous Vehicle Liability Gap Leaves Crash Responsibility Unresolved
Autonomous transport systems operating without human control create an unresolved legal void over liability when incidents occur. Governments and operators face regulatory and financial exposure until clear accountability frameworks are legislated and enforced.
AI in Elder and Child Care Raises Manipulation and Privacy Governance Risks
AI systems deployed in elder and child care carry documented risks of psychological manipulation and clinical misjudgement, while AI-driven medical research exposes patient data to inadequately governed privacy risks. Boards face mounting regulatory and reputational liability without robust data governance frameworks and patient rights protections in place.
AI-Driven Market Monopolisation Through Algorithmic Price Control
AI systems controlling pricing mechanisms enable firms to abuse market power and suppress competition through algorithmic coordination. Boards face regulatory scrutiny and reputational risk where automated pricing strategies breach competition law.
Superintelligent AI Agents Risk Becoming Uncontrollable as Autonomy Increases
Advanced AI agents operating at high autonomy levels may exceed human capacity to oversee or intervene, a problem research indicates has no reliable technical solution. Governments and boards that deploy autonomous AI systems without enforceable control mechanisms face irreversible loss of operational oversight.
AI Agent Decisions Cannot Be Reliably Predicted Across All Situations
AI-based agents exhibit decision unpredictability that prevents operators from anticipating system behaviour under novel or edge-case conditions. Boards cannot assure regulators or insurers of safe outcomes where agent actions remain opaque and unforeseeable.
AI Decision Opacity Blocks External Accountability in Public-Sector Systems
AI models in government ecosystems cannot reliably show external parties which data inputs drove a given decision. This opacity undermines regulatory scrutiny, legal challenge rights, and public trust in automated public-sector processes.
AI Systems Causing Human Harm Through Unsafe Agent Actions
Learning models can harm humans both directly and indirectly, and existing safety frameworks derived from Asimov's laws remain insufficient to constrain autonomous agent behaviour reliably. Governments face material liability and public trust risk if deployed AI systems lack robust, legally grounded safety assurance mechanisms.
LLMs Accelerating Dual-Use AI Development at Scale
Large language models can autonomously build new AI systems and adapt existing ones for high-risk applications, compressing development timelines. Boards face material liability exposure as dual-use capability proliferation outpaces regulatory oversight and internal governance controls.
AI Model Reproducibility Failures Undermine Scientific and Regulatory Trust
AI models trained on varied datasets with large parameter spaces routinely cannot be reproduced, making independent validation impossible. Regulators and procurement bodies cannot verify claimed performance, exposing public institutions to unauditable algorithmic decision-making.
AI Systems Exposing Personal Data Through Inadequately Secured Channels
AI deployments create exploitable vulnerabilities through which personal information can be accessed without authorisation. Boards face regulatory exposure and loss of user trust where data governance frameworks fail to address AI-specific privacy risks.
Generative AI Systems Amplify Societal Bias and Suppress Output Diversity
Generative AI models trained on non-representative data reproduce and intensify historical biases, creating measurable performance disparities across demographic groups and languages. Organisations face legal exposure, reputational harm, and flawed decision-making where homogenised outputs go unchallenged in operational processes.
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