How our content is made

Every article is an original write-up produced by our automated editorial pipeline built on large language models (currently Anthropic's Claude models). From publicly reported, factual news it produces a standalone piece in Polish (and then an English translation). We do not copy anyone's material verbatim — we publish our own write-ups of the facts, and cite our sources at the end of each article.

We say this openly because readers deserve to know how the text they are reading came to be. It is also required by Article 50(4) of the EU AI Act: AI-generated or substantially AI-modified text published to inform the public on matters of public interest must be disclosed.

Since August 2026 no new article reaches the site automatically. The process works like this:

• the system drafts an article and records the date and time it was created,
• an editor reads the full text and checks the sources — the approval form requires confirming the whole article was read; rubber-stamping is not possible,
• the editor may make corrections; every edit is recorded (date and a description of the change), and after approval the text is never modified by AI again,
• only after approval is the article published, with its full history visible under the headline (“Publication history”).

Editorial responsibility for every published piece rests with a human — the site's founder and editor, Patryk Raba.

• “Prepared with AI tools, verified by our editorial team before publication” — articles published under the current process, with a full publication history (created, approved, edits),
• “Published before our current editorial review process launched” — older archive articles (before August 2026); they were also AI-assisted,
• the “nowosci.ai newsroom”byline — AI-assisted text is never signed with a person's name as its author; the person responsible for verification is credited separately.

• we never sign AI text with a human name as its author,
• we do not generate photorealistic images of real people or events — photos come only from licensed sources (see below),
• we do not publish articles without citing sources,
• we do not modify an article with AI tools after an editor has approved it,
• we do not delete publication history — the created / approved / edits record is permanent.

We correct errors reported to redakcja@nowosci.ai; significant corrections are noted in the article's publication history.

The “Radio” feature and single-article playback are machine-generated narration (text-to-speech) of our own article text. We use the Piper engine (open-source software, voice pl_PL-darkman-medium) and, as a secondary option, the ElevenLabs service. Every recording is rendered and hosted on our own infrastructure (assets.raba.pl).

The apps play no third-party audio whatsoever: there is no music, no podcasts, no radio stations, and no streams from external services. The entire audio track is speech synthesis of our own text.

Illustrations come only from lawful sources, and we keep attribution in an internal register (image-credits.json):

• a licensed Adobe Stock pool (commercial license),
• official press / newsroom galleries for editorial use (e.g. Microsoft, Google, NVIDIA) with the required source credit,
• Wikimedia Commons under free licenses only (CC BY), whose type we verify programmatically (rejecting “NC”, “ND” and “all rights reserved”).

All files are self-hosted, under nowosci.ai/images.

The Nowości AI apps fetch a single, first-party data feed (nowosci.ai/api/feed) and display only our content. They contain no search or catalog of other parties' media, connect to no external streaming services, and provide no access to third-party content.

We hold all rights or permissions necessary to publish the text, audio and images shown in the service and the apps. For any copyright or licensing matters, contact redakcja@nowosci.ai.