Skip to content

AI policy

AI Pulse stories are written by artificial intelligence, checked automatically, then reviewed one by one by a member of the editorial team, who decides whether to publish them. Every story says so. Here is how they are produced and checked.

The steps

  1. Monitoring. Official feeds (AI labs, companies, institutions, research publications) are checked several times an hour, within their terms of access.
  2. Clustering. Items about the same event are grouped. If we already covered the event, the existing story is updated instead of publishing a new one.
  3. Selection. A model scores the topic (novelty, impact, strength of sources) and picks a format: brief, story, explainer.
  4. Research. Primary sources are retrieved and broken down into dated facts, each linked to its source. Without a primary source, or two independent ones, the topic is dropped, unless a member of the team keeps a story broken by a single reputable outlet (see our editorial standards).
  5. Writing. A model writes the story from those facts only, following our editorial standards.
  6. Fact-checking. Another model checks every claim against the facts and sources: names, figures, dates, quotes, versions. Anything unsupported is removed or attributed.
  7. Quality control. A story is published only if it passes every check: fact-check score, no reuse of source text, no duplicate of our own stories, complete structure, real added value over the sources. Otherwise it is revised, then rejected.
  8. Review. A member of the editorial team reads every story that passed the checks and decides to publish it, send it back or drop it. Nothing is published without this approval, and a story can be withdrawn at any time.

The tools

Selection, research, writing and fact-checking use Anthropic's Claude models, and its web search tool to find an original document. Topic clustering runs on our own servers, by comparing words, with no external service. These services process public texts; no reader data is sent to them.

What AI does not do

Transparency

Every story is labelled “written with the help of artificial intelligence”. Article 50 of the EU AI Act (Regulation (EU) 2024/1689) requires AI-generated text published to inform the public to be disclosed, unless it has been reviewed by a person who holds editorial responsibility for it. Our stories are reviewed, but we keep the label for transparency. Found an error? Report it (see contact): corrections are listed on our corrections page.