AIBlindspot

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.

Explore

1296 cases

Lifecycle quick filter:DesignDevelopDeployOperate
DATDAT-0023/5GovernmentGlobal

Generative AI Systems Reproducing Private and Copyrighted Data

Generative AI models trained on unlawfully collected data risk reproducing personally identifiable information, medical records, and copyrighted content. Government bodies face regulatory liability and reputational harm where procurement and deployment lack robust data governance controls.

Source: MIT AI Risk Repository — Evaluating the Social Impact of Generative AI Systems in Systems and Society (Solaiman2023)Ingested —
DATDAT-0034/5GovernmentGlobal

Generative AI Systems Deliver Unequal Accuracy Across Language and Demographic Groups

Generative AI systems trained on English-dominated internet data systematically underperform for non-English speakers, minority language groups, and racially distinct speech patterns. Government deployment of such systems risks embedding structural inequality into public services and exposes departments to legal and reputational liability.

Source: MIT AI Risk Repository — Evaluating the Social Impact of Generative AI Systems in Systems and Society (Solaiman2023)Ingested —
DATDAT-0014/5OtherGlobal

Generative AI Systems Embed Culturally Contingent Values, Creating Global Deployment Risk

Generative AI cannot be culturally neutral; definitions of harmful content vary by region, language, and political context, making a universal safety standard unattainable. Organisations deploying models globally face material liability and reputational risk where outputs deemed acceptable in one jurisdiction are unlawful or offensive in another.

Source: MIT AI Risk Repository — Evaluating the Social Impact of Generative AI Systems in Systems and Society (Solaiman2023)Ingested —
DATDAT-0034/5TechnologyGlobal

Generative AI Systems Embed and Amplify Bias Against Marginalised Groups

Generative AI amplifies harmful biases across the full machine learning pipeline, including modelling choices, compression, and hardware, not data alone. Boards face legal, reputational, and regulatory exposure where AI products cause representational harm to protected or marginalised groups.

Source: MIT AI Risk Repository — Evaluating the Social Impact of Generative AI Systems in Systems and Society (Solaiman2023)Ingested —

Beyond accidental failureNational Security

We also track 20 hostile uses of AI.

National Security dashboard →

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