AI against modern slavery (AIMS)

Author

In brief

Around 50 million people live in modern slavery, most of them exploited inside private-sector supply chains. Modern Slavery Acts in the United Kingdom, Australia and Canada oblige thousands of companies to disclose their countermeasures each year, yet almost none of those statements are ever read at scale. Project AIMS builds the open research foundation that scalable assessment would need: the largest annotated dataset of modern slavery statements, fine-tuned language models, explainable cross-jurisdictional review, distillation for low-resource deployment, and a quality-assessment framework that reaches beyond bare legal compliance.

“The transparency mandate exists. The data infrastructure to deliver on it does not.”

What the article covers

• AIMS.au: 5,731 statements filed under Australia's Modern Slavery Act, annotated at sentence level against eleven labelling questions derived from the Act's seven mandatory criteria.
• AIMSCheck: compliance review with explainability built in from the start, including SHAP token attribution and evidence-status tracking.
• AIMSDistill: a 340-million-parameter student model that runs roughly seven times faster than the teacher ensemble it replaces and uses about six times less energy.
• AIMS-QA: thirteen benchmarking methodologies synthesised into a single taxonomy, assessed by a two-agent pipeline with a human reviewer always in the loop.
• A design brief for legislators: structured submission formats, persistent corporate identifiers, public APIs, and criteria drafted with assessability in mind.

How to cite

Adriana Eufrosina Bora. “AI against modern slavery (AIMS).” vISIon: The ISI Magazine Vol. 1 No. 2, June 2026, pp. 3–8.