BREAKING LIVE All breaking →

UCAPITAL NEWS

Editorial and AI policy

Part of the stories on UCapital News is written by UCapital AI desks: artificial intelligence systems that work on market data and verifiable sources, under fixed rules and automated checks. This page explains what they do, where their information comes from, how they are controlled and how to report an error.

DRAFT pending legal review: the wording may change before final publication.

What the AI desks are

An AI desk is an automated newsroom dedicated to one news area, not a person. Every story it writes is signed «UCapital AI Desk» with the desk name and carries the ✦ AI mark. Two desks are active today:

  • UCapital AI Desk · Global Markets & Macro, in English: global markets, macroeconomic data, central banks.
  • UCapital AI Desk · Mercati & Macro Italia, in Italian: the Milan and European market open and close, macroeconomic data, central banks.

The English and Italian versions of the same story are each written by their own desk, from their own sources: they are not translations of each other.

Where the data and the news come from

AI desks only write from information retrieved while they prepare the story:

  • market data from the UCapital platform (prices, indices, economic and corporate calendars), supplied by financial data providers;
  • news from agencies, publishers and third-party providers, and official announcements from issuers and institutions.

Every story lists its sources in the «Sources» section, linking to the original. Reports built on market data only cite «UCapital market data» as their source, linking to the instrument page on the platform.

Rules and checks

Before publication every story goes through automated checks that do not depend on the AI model:

  • No number without a source: every figure, percentage or price must match retrieved data or a cited source. A text with figures that cannot be traced to a source is never published automatically.
  • No recommendations: AI desks do not advise buying, selling or holding financial instruments, give no price targets and use no promotional language. Impact indicators, where shown, describe the possible effect of a story on the markets, not an invitation to trade.
  • No invented sources: links and sources may only point to material actually retrieved.
  • No copying: a story may not rewrite the text of a single source, it must add data and context.

Risk tiers and human review

Every story gets a risk tier, computed deterministically by the rules and never lowered by the model:

  • T0: published directly; applies to reports built on market data only.
  • T1: published and queued for a newsroom check within two hours.
  • T2: held until a human reviewer in the newsroom approves it; if rejected it is not published. Any flag raised by the automated checks puts the story at this tier at least.

During the launch phase every AI desk story is treated as T2 and goes through human review. Every publication is logged with the sources used, the tier assigned and the reviewer, if any.

How to report an error

If you find a wrong figure, an irrelevant source or a text that looks incorrect, write to the newsroom at the address below, with the link to the story and what is wrong. Every report is read by a person.

support@ucapital.com

Corrections

When an error is confirmed we correct the story at the same address, and the update date appears in the byline. If the error changes the meaning of the story, the story is withdrawn. A desk that keeps making errors is suspended until the cause is fixed: every desk can be stopped at any time.

Nature of the content

UCapital News content is for information only and is not financial advice, a recommendation or a solicitation to invest. AI can make mistakes: always check the original sources before making any decision.

AI use policy. UCapital AI supports the newsroom with news analysis, summaries, market reactions, translations and the generation of cards and charts. AI-generated or AI-assisted content is labelled in the byline (AI Desk or joint byline) and is reviewed by editors before publication. News received via API feeds from third-party providers is attributed to the original source. AI can make mistakes: report any inaccuracies to the newsroom. Users’ personal data is not used to train the models.