retronewsreaderBack to the reader

AI use

Last changed 8 October 2026
Questions? Contact reader@retronewsreader.com

A note from the developer on where artificial intelligence is at work while you use the reader. There is one such service: Jev, from TypeSafe AI. It is a classifier. It is given a question and a fixed list of possible answers, and it picks one. It writes nothing you read.

One more service is used behind the scenes, for the person who runs the reader only: Claude, from Anthropic, writes the summaries in their evening brief (the end of this page). No reader gets it.

What it does

What it is sent

What it does not do

The evening brief (a trial, for the person who runs the reader)

The person who runs this reader can turn on one email a day that summarises the stories they saved themselves. The summaries are written by Claude, an AI model from Anthropic. For each story it is sent the outlet's name, the headline, the byline, the date and the story's text: read from the outlet's page by the reader's server where that page can be read, otherwise taken from the feed. A page for subscribers is not read. It is told nothing about any reader: not who is asking, and not what anyone else has saved or read. The summaries go into a PDF that only that account can open, kept for 30 days. Readers do not get it, and nothing it writes appears in the reader. What Anthropic does with what it is sent is in its terms and privacy policy.

Where it can be wrong

Topic and place labels are a best guess and are sometimes wrong. A mislabelled story is still in your list: it is only filed under the wrong tag. If the service cannot be reached, stories arrive without labels and everything else works as usual.

What TypeSafe AI does with what it is sent is in its own terms and privacy policy.

What the reader keeps about you is on the privacy page.