AI policy you can trace back to the source

AI regulation is now global, but the tools for following it are uneven. Large jurisdictions are covered by many trackers; South Asia, ASEAN, Africa, the Gulf and Latin America are often reduced to a paragraph. Summaries circulate without links to the instrument they describe, and readers cannot tell whether a date was checked last week or two years ago.

This project takes a narrower position. Every record cites an official source and states when it was last checked. Regulatory status uses a small, explicit vocabulary. Obligations are broken out so they can be mapped to controls and evidence. The whole dataset is open, versioned and reviewable in public.

By jurisdiction

Start from a country or bloc: regulatory status, binding rules versus guidance, regulators and official sources.

By instrument

Filter acts, regulations, strategies and guidance by status, actor, sector and use case, then open the record for dates, scope and obligations.

By obligation

Read requirements as practical tasks with the article they come from, who they bind and what evidence they imply.

By risk

Move from a category of harm to the incidents recorded for it and the policies that respond, using the MIT AI Risk Repository taxonomy as a shared vocabulary.

By date

Follow the change log and RSS feed for dated, source-linked developments and upcoming application dates.

As data

Download the dataset, call the read-only API or point an AI assistant at llms.txt; everything carries its source and licence.

1. A record is created from an official document (legislation, gazette, regulator or ministry publication) with its URL, publisher and access date.
2. Status, dates, scope and obligations are written in plain language with article references; no legal commentary is copied.
3. A reviewer opens the source, confirms each field and sets the verification date. Until then the record is shown as source-linked, not verified.
4. Changes are logged with what changed and its practical impact, and the dataset is versioned in the open repository.

Full methodology, status definitions and source tiers

The project stands on open work by others. Each is credited where it is used, and the licences below govern reuse of that material.

1. Slattery, P., Saeri, A. K., Grundy, E. A. C., Graham, J., Noetel, M., Uuk, R., Dao, J., Pour, S., Casper, S., & Thompson, N. (2025). The AI Risk Repository: A comprehensive meta-review, database, and taxonomy of risks from artificial intelligence. MIT AI Risk Initiative, arXiv:2408.12622 · CC BY 4.0
2. McGregor, S. (2021). Preventing Repeated Real World AI Failures by Cataloging Incidents: The AI Incident Database. Proceedings of the AAAI Conference on Artificial Intelligence (IAAI-21); Responsible AI Collaborative · CC BY-SA 4.0
3. Weidinger, L. et al. (2021). Ethical and social risks of harm from Language Models. DeepMind, arXiv:2112.04359
4. National Institute of Standards and Technology (2023). AI Risk Management Framework (AI RMF 1.0). NIST · Public domain (US Government work)
5. European Parliament and Council (2024). Regulation (EU) 2024/1689 (Artificial Intelligence Act). Official Journal of the European Union · EU reuse policy (Decision 2011/833/EU)