Open data

Take the data. Keep the credit line.

One endpoint, no key, no sign-up. It returns the AI Act event record as JSON or CSV, up to 1000 events per call, with the primary source URL on every row. Published under CC BY 4.0, so you can use it in research, in a product, or in a client report as long as the attribution stays on.

Copy and run

Both calls work from any terminal, unauthenticated. Add ?since=YYYY-MM-DD to cut the window and ?source= to pick one source.

JSON, since a date 
curl -s "https://api.aiactradar.com/v1/public/export?since=2026-01-01" \
  -H "Accept: application/json"
CSV, one source, to a file 
curl -sL "https://api.aiactradar.com/v1/public/export?format=csv&source=eu_ai_office" \
  -o aiactradar-events.csv

format is json by default and accepts csv. CSV is served with a download filename, RFC-4180 quoting, and the licence line in the first two rows so it survives being emailed around. The source list lives on the sources page.

What you get
id
Stable event identifier. Never reused.
source_id
Which official source the entry came from.
event_type
guidance, standard, consultation, enforcement, national, court, meta.
severity
info, low, medium, high, critical — our reading of how much it moves.
title_en
English title.
title_de
German title where the source or our pipeline provides one, otherwise empty.
summary_en
Short English summary of what changed.
url_official
The primary source document. Every entry has one.
ai_act_articles
AI Act articles the entry touches.
topics
Topic tags, e.g. gpai, transparency, standards.
published_at
Publication time at the source. Epoch milliseconds in JSON, ISO-8601 in CSV.
Licence

CC BY 4.0

Copy it, change it, build on it, including commercially. The one condition is naming the source. Use this line:

AI Act Radar (aiactradar.com), CC BY 4.0

Full terms: creativecommons.org/licenses/by/4.0 . The same line ships in every response, so it travels with the file rather than living only on this page.

Snapshot, not a stream

For updates, don't poll this.

The export is a point-in-time snapshot capped at 1000 events. It is built for a one-off pull: a dataset, a backfill, an analysis. If you need to know when something changes, take the push path instead.

  • Signed webhook — every new event delivered to your endpoint, HMAC-signed
  • MCP server — your agent queries the live record directly
  • Public RSS feed — the last 50 events, no key
Get free access
On data quality

What we will and won't claim.

Every entry carries the primary source document it came from, so you can check any single row against the original rather than trusting our summary of it. Titles, dates and URLs come straight from the source. The classification layer on top — severity, topics, and the AI Act articles an entry touches — is our reading, produced by a pipeline with human review, and where it is uncertain the entry says so instead of guessing confidently. Coverage gaps are named on the sources page, including the adapters that are currently switched off. Informational content, not legal advice: for a decision that matters, read the linked source and talk to counsel.