How we use artificial intelligence
Last revised: internal draft, November 2026.
Where Foro uses artificial intelligence
| Use | What the AI does | Human review |
|---|---|---|
| Company profiles | Drafts a structured summary (business, figures, capital structure, recent milestones) from public documents supplied to it as the source. | The Foro team reviews each profile before publishing it. The company can validate its figures or ask for corrections. |
| Answers to questions | Drafts an answer to a user's question, citing specific extracts from public documents as the only permitted source. | The Foro team reviews each answer before publishing it. If the company is on Foro, it can first approve it, complete it or propose its own answer, which also goes through Foro. |
| Lumia, the assistant | Answers what you ask about a company in a private conversation, using only the figures on its profile and their sources. | It is not published: it is an answer for you. Every figure links to its document so you can check it. |
| Question admission filter | Automatically classifies every question sent and instantly rejects those asking for a directional view on the price, a price target, a valuation, or unpublished information. | Automatic at the moment of sending. It needs no prior review because it generates no content: it only admits or rejects. |
| Podcast scripts and voices | Drafts the script from profiles and documents already approved, and generates the audio with synthetic voices. | A Foro editor approves the script and then the final audio before publication. |
What Foro's AI never does
- It never drafts content that does not cite a specific, verifiable public source.
- It never produces views on whether to buy, sell or hold a security, or price targets.
- Nothing is published without the Foro team having reviewed it.
- It never uses personal user data to generate editorial content: profiles and answers are built only from the company's own documents.
The provenance dossier
Every profile and every answer carries a provenance dossier: the list of specific statements in the text together with the public document and date they come from. You can read it on each profile. If a statement cannot be anchored to a quotable extract, the system is designed to leave it out rather than invent it.
Synthetic voices
Foro's podcast episodes are narrated by AI-generated voices (provider: ElevenLabs), not by human presenters. Each episode says so audibly at the start and visibly on its player.
Which model providers we use
Foro uses Anthropic language models to draft profiles, scripts and answers, and ElevenLabs speech synthesis models for episode audio. Neither provider receives personal data about Foro users: they only receive the company's public documents and, for questions, the text of the question sent (with nothing identifying the person who asked it).
How to report an error
If you spot an incorrect statement, a misattributed quote, or any other fault in AI-generated content, write to us at [PENDING: editorial contact email]. Every correction is reviewed and, where appropriate, published along with the date it was corrected.
This is a translation provided for convenience. The Spanish version is the one that governs.