AI supplier change log

Nymbrink follows the public pages of 19 AI and software suppliers: 18 privacy policies, 5 subprocessor lists, 10 usage policies and 20 sets of terms. It records each change in their wording exactly as it was read. It quotes; it does not judge.

RSS feed

Latest changes

Suppliers and pages

Nymbrink follows only the documents listed under each supplier. A change to any other document would not appear here, and neither would a change made before the first reading or record shown.

How this log works

For the pages Nymbrink reads itself, it reads each once a day as a plain visitor, following the site’s rules for automated readers. A change is listed only once two checks in a row have seen the same new wording.

Some suppliers’ sites turn automated readers away. For those, entries show the documents as Open Terms Archive recorded them (GenAI Governance Archive, University of Bremen). A recorded change is listed once the archive has recorded nothing else in that document for seven days, and the archive’s own reformatting of its copies is never counted as a change. Nymbrink did not read those pages itself and has not signed those entries.

This shows which wording changed, quoted exactly from the recorded versions of the public page. It does not say whether a change is better or worse, why it was made, or what it means for anyone. The supplier’s own page is the authority on what it says.

This log covers the wording of the pages listed here. It does not read incident announcements on status pages, blogs or social media, and it does not detect incidents. An incident notice reaches Nymbrink only when a supplier with a passport publishes a signed one.

Each of Nymbrink’s own daily readings is kept in a record Nymbrink signs, which anyone can check with the open Provenance tools. If an entry is wrong, write to contact@nymbrink.com.

What Nymbrink cannot read itself

  • OpenAI, Perplexity and xAI: their sites turn automated readers away, and Nymbrink does not work around that. Their privacy policies are shown as Open Terms Archive recorded them.
  • Meta: its pages are put together by scripts after loading, which Nymbrink’s reader does not run yet. Its EU AI terms are shown as Open Terms Archive recorded them; its privacy policy is not covered.
  • Notion: its pages are put together by scripts after loading, which Nymbrink’s reader does not run yet.
  • Anthropic, Slack and Adobe: Nymbrink’s reader does not get the full text of Anthropic’s subprocessor list, Slack’s main services agreement and acceptable use policy, or Adobe’s general terms, so those documents are not followed.
  • Canva: its terms pages turn automated readers away, and Nymbrink does not work around that. Its privacy policy is followed.
  • Google: its Gemini API terms came back in different languages between readings, which would look like changes that never happened, so they are not followed.

Entries marked “Recorded by Open Terms Archive” are derived from the GenAI Governance Archive, University of Bremen, by Open Terms Archive contributors, licensed under the Open Data Commons Attribution License 1.0.