How we use artificial intelligence

1In brief

In the app Sinapsi’s artificial intelligence has a name and a figure: it is called Calypso, it is drawn as a chess rook and it always introduces itself as an artificial intelligence.

This page is the information that Article 13 of Italian Law no. 132 of 23 September 2025 requires of professionals who use artificial intelligence, also written in line with the transparency obligations of Regulation (EU) 2024/1689, the AI Act.

2What artificial intelligence does

• Reads invoices, bank movements, financial statements, contracts, credit data and legal documents, including from photos or PDFs.
• Links documents together, for example an invoice with its payment.
• Calculates indicators, deadlines and trends.
• Forecasts, in the Vedi prima window, where you will close the month, quarter and year, the figures for the next three months and upcoming events, with a range and a probability.
• Proposes concrete moves, with their estimated value in euros.
• In litigation it reads the document, calculates the deadlines, estimates the effects and proposes a line.
• Drafts summaries, notes and replies, which a person rereads.
• Warns and explains in the app in Calypso’s voice: the morning summary, warnings, answers to questions and the missions of the tower game.

3What people do

Every day the Celebron team checks data quality and verifies proposals before they appear as checked: next to each one is the role of the person who verified it. Litigation lines are assessed and signed by a qualified professional: a lawyer, accountant or tax adviser. The intellectual work remains human; AI is a support tool, as the law requires.

4What artificial intelligence does not do

• It does not decide for you: decisions remain with the business owner.
• It does not grant or refuse credit and assigns no credit scores.
• It provides no legal advice and does not defend in court.
• It gives no investment advice.
• It does not handle criminal proceedings.
• It does not use your data to train third-party models.

5How far you can rely on it

Forecasts are estimates, not guarantees. That is why we show them with a range and a probability, not as a single figure. Their quality depends on how complete and up to date the documents are and on future events that nobody controls. AI can make mistakes, for example by misreading a document or overestimating an effect: human checks are there to catch them, and over time we compare forecasts with actual results to make them more reliable.

If a figure or a proposal looks wrong to you, tell us: a person will review it.

6Which systems we use

We use general-purpose AI models from leading providers, through professional services that exclude the use of customer data to train the models, and Celebron’s own calculation models for forecasts, deadlines and indicators. The providers are listed in the Data Processing Agreement.

Celebron has assessed Sinapsi under Article 6 and Annex III of Regulation (EU) 2024/1689 (AI Act): Sinapsi is not a high-risk artificial intelligence system. As designed, it:

• does not assess the creditworthiness of natural persons and assigns no credit scores, even when the Client is a sole proprietorship: it shows the business its own data, without judgements;
• is not intended for banks or lenders to decide whether to grant credit;
• does not evaluate workers and takes no decisions on hiring, tasks, promotion or dismissal: in the Dashboard, employee data appear only as totals;
• is not used by or on behalf of judicial authorities: the litigation folder prepares proposals for the Client and its professional.

The assessment is documented and repeated before any new feature that could change it. The transparency obligations of the Regulation continue to apply. As a precaution we already apply to Sinapsi the main safeguards for high-risk systems and have a plan in case the classification changes: the binding rules are in the Artificial Intelligence Terms.

7Your data and AI

The management, control and custody of the data are Celebron’s responsibility: protected servers of an official cloud partner, in the European Union, with the safeguards of the Data Processing Agreement. AI receives only the data needed for the task. We take no decisions based solely on automated processing that produce legal effects on individuals.

8How to recognise what AI does

In the Dashboard, forecasts and proposals produced by AI are marked as such, with the range, the probability and, once verified, the role of the person who checked them. Calypso’s messages always carry its name and figure; when a person on the Celebron team replies, their name and role are shown. The tower game shows only how complete and up to date the data are: it is not a judgement on the business or a credit score. In the demo dashboard on celebron.it the numbers and cases are fictional, generated to show the service.

9Who is accountable

At Celebron the use of AI has a designated manager, who looks after its rules, team training and periodic checks. The people working on Sinapsi are trained in the proper use of AI and its risks. For questions, reports and requests for review by a person: sinapsi@celebron.it or info@celebron.it.

10References

• Regulation (EU) 2024/1689 on artificial intelligence (AI Act), as amended.
• Italian Law no. 132 of 23 September 2025 on artificial intelligence.
• Regulation (EU) 2016/679 (GDPR) and Italian Legislative Decree no. 196 of 30 June 2003.