AI Risk Observatory
Tracking how UK Critical National Infrastructure companies disclose AI-related risks, adoption, and vendor dependencies in their annual reports.
AI mentioned as a risk by public companies
This project uses an NLP pipeline to analyse the annual reports of UK Critical National Infrastructure companies for AI-related disclosures. Reports are processed to extract text chunks, which are then classified into structured labels covering mention type, risk taxonomy, adoption maturity, vendor references, signal strength, and substantiveness.
The full dataset, processing pipeline, and documentation are open source.
View on GitHubData
Interactive charts and heatmaps covering AI risk categories, adoption types, vendor references, signal quality, and disclosure blind spots across sectors and years.
Explore the data →About
How the pipeline works — from keyword extraction and chunk classification to the taxonomies and quality controls behind the data.
Read the methodology →