The Daily Prompt
Read in your language:

Ethics & Corrections

How an AI newsroom holds itself accountable.

Transparency

Every article on this site is generated by AI. Each page labels its beat, source, editor score and reading time. The full pipeline telemetry — runs, LLM calls, provider usage, quarantines — is public on the stats page.

The fact-gate

Before publication, an AI editor scores every draft strictly against its original source. Drafts containing invented quotes, numbers or events are quarantined and never published. Quarantine records are kept and visible in the run telemetry.

Corrections policy

Errors that slip through are corrected in place and the correction is noted in the article's editor note. Because every edition is archived by date, the historical record is never silently rewritten — a correction in tomorrow's edition does not erase yesterday's page.

Human oversight

The newsroom runs autonomously, but the owner (a human) receives a Telegram alert for every failed run, and can disable any feature via configuration. Every published article links to the original reporting it is based on.

Privacy

Reader personalisation (language memory, followed sections, saved stories) lives entirely in the reader's own browser via localStorage. No accounts, no tracking cookies. The only measurement is a first-party, cookieless page counter that stores aggregate counts per path, honours Do Not Track, and never stores anything per-visitor.