Skip to content
ModelVault
Home Services Demo Contact Sign in
IT Italiano EN English FR Français ES Español DE Deutsch JA 日本語 KO 한국어
Legal

Terms & Conditions

Last updated: 01/10/2026

1. Purpose

These Terms govern access to and use of modelvault.it and the services provided by Federico Sabbadini through ModelVault, including the development of custom AI solutions and the related subscription service.

2. Service Structure

The ModelVault service consists of development, configuration, hosting and controlled access to AI panels/APIs. Access is governed by a plan configured in the ModelVault control database and may include limits on API calls, documents processed, compute time, tokens, benchmarks or other agreed operational metrics.

3. Duration, Renewal and Payment

The contract has a minimum duration of 12 months from the service activation date. On expiry it automatically renews for further 12-month periods, unless cancelled in writing with at least 30 days' notice.

Payment may be monthly or annual in advance. Annual payment includes a discount agreed at the contract stage. At the end of each month of use, ModelVault performs a true-up between the prepaid amount and actual usage: any excess is charged, any credit is deducted from the following month.

Before the contract is activated, the client is entitled to a free trial period (typically 7 days) during which ModelVault monitors actual usage costs. At the end of the trial, ModelVault puts forward a pricing proposal based on actual consumption data.

4. Delivery and Access

Upon completion of development, the client receives credentials and/or API keys for the contracted services. Credentials are personal, non-transferable and revocable. The user is responsible for their safekeeping and for any use made through their account or keys.

When access takes place via the official website, the panel, or ModelVault's server-side proxy, ModelVault controls the application perimeter and applies its own authentication, authorisation, logging, usage-tracking and security measures.

When, instead, API keys are provided for use through the client's own sites, front ends, back ends, scripts, apps, plugins or external interfaces, safekeeping and protection of the keys is the client's responsibility. Requests made with a valid API key are considered authorised client use until the key is suspended, rotated or revoked, and may be counted towards plan limits, overages and the final invoice even in the event of exposure, misuse, loss or theft on the client's side.

CORS allowlists, where configured, are an additional browser-level restriction and do not replace API key security, server-side authorisation, or the client's responsibility for protecting credentials.

5. Usage-Based Billing and True-Up

Billing is based on the actual usage of each service. There are no predefined fixed plans: each service has different variable costs that are automatically tracked by the metering system.

The cost components tracked for each project may include, depending on the service:

  • CPU time — compute time consumed per request
  • LLM tokens (input and output) — for services using language models via external providers
  • Web searches — for services that query external search APIs
  • Documents and pages processed — for document-extraction services
  • Embedding operations — for semantic search services
  • Storage — disk space used by indexes, vectors or persistent output
  • Latency and request count — response time and number of API calls per client

Infrastructure costs (VPS, maintenance, backups) are not included in the net per-project costs and are assessed separately by the operator. The client sees only the usage and the charged cost agreed under the contract.

If the agreed limits are exceeded, the service is NOT blocked. ModelVault internally flags the overage and agrees with the client how to handle it (true-up, plan adjustment, or contract review).

Metrics may include API calls, response time, compute time, tokens, documents processed, specific benchmarks, or indicators collected via reverse proxy, application back ends, Prometheus, or other monitoring systems.

ModelVault may collect and retain technical metering data necessary for security, abuse prevention, limit enforcement, billing, overage calculation, provider cost allocation, debugging, auditing and capacity planning. This data may include request identifiers, timestamps, source and destination, service and endpoint called, HTTP method and status, latency, bytes transferred, API key hash or client label, origin or integration used, tokens, documents, pages, model or provider, estimated CPU/GPU/RAM/storage usage, and estimated internal, provider or billable costs. Raw API keys are not intentionally stored in metering logs.

Default retention, unless otherwise agreed in writing: raw events for 90 days, summaries for 730 days, audit/billing records for 1095 days. Specific contracts may specify different retention periods.

6. Use of AI Services

AI models are support tools. Their output does not constitute professional advice (legal, medical, financial). The user is responsible for the use and supervision of results, especially in healthcare or regulated settings.

7. Free Access, Demos and Trials

Free access, demos, trials, public previews or academic access are granted at ModelVault's discretion. ModelVault reserves the right to limit, suspend or close them at any time, with no obligation of prior notice, refund, compensation or continuity of service.

8. Client Data

Data provided by the client for model training remains the client's property. ModelVault uses it exclusively to deliver the commissioned service. Upon termination of the contract, the data is deleted within 30 days, unless otherwise agreed in writing.

9. Intellectual Property

Models, code, architecture and design belonging to ModelVault are the exclusive property of Federico Sabbadini. Clients acquire a usage licence limited to the contracted services for the duration of the contract, not a right of ownership or sublicensing.

10. Suspension and Termination

ModelVault may suspend or revoke access to panels, APIs or services in the event of non-payment, contract expiry, exceeded limits, abuse, security risk, unauthorised use, or operational necessity. Suspension may be applied even at the level of a single API or a single service.

Reactivation may require payment of amounts due, contract renewal, selection of a new plan, or new commercial authorisation.

11. Limitation of Liability

ModelVault does not guarantee continuous service availability and is not liable for direct or indirect damages arising from use or inability to use the service, except in cases of wilful misconduct or gross negligence. Unless otherwise agreed in writing, uptime, latency, model accuracy, throughput or continuity of access are not guaranteed.

12. Governing Law

These Terms are governed by Italian law. Any dispute falls under the exclusive jurisdiction of the Court of Brescia (Foro di Brescia).

13. Contact

Legal communications: info@pec.modelvault.it (PEC, certified email)
Support: info@modelvault.it

© 2026 ModelVault — Federico Sabbadini
Privacy Policy Cookie Policy Terms & Conditions