Terms of Service — OpenReef
Last updated: 12 July 2026
Operator: OpenReef (“we”, “us”, “our”)
Service: https://openreef.network and related APIs (the “Service”)
Contact: via Telegram or the contact channel published on the website
**Disclaimer:** This document is a practical operating draft for product launch and hosting compliance. It is **not** a substitute for advice from a qualified lawyer in your jurisdiction. Have it reviewed before relying on it for regulated activity or large-scale EU processing.
1. Acceptance
By creating an account, accessing, or using the Service, you agree to these Terms and to our Privacy Policy. If you do not agree, do not use the Service.
If you use the Service on behalf of an organisation, you represent that you have authority to bind that organisation.
2. The Service
OpenReef provides a web platform to:
- upload training datasets (e.g. JSONL, CSV, TXT);
- select or import base models (including from Hugging Face, subject to their terms and licences);
- configure fine-tuning jobs (LoRA / QLoRA presets);
- run jobs via mock, local, or decentralised OpenGPU Network compute adapters;
- pay with credits (including Stripe where enabled) and download trained adapters.
The Service is provided on a commercial / beta basis as described on the website. Features may change.
3. Accounts
You must provide accurate registration information and keep credentials confidential. You are responsible for activity under your account.
We may suspend or terminate accounts that violate these Terms, applicable law, or that pose security or abuse risk.
4. Credits, payments, and refunds
- Jobs may require USD credits purchased via Stripe or allocated in development/test modes.
- Prices shown before confirm are estimates; final settlement follows our ledger rules (including partial refunds on cancel/failure by phase).
- Except where mandatory law requires otherwise, credits are non-transferable and may expire only if we publish an expiry policy in advance.
- Chargebacks or payment fraud may result in suspension.
5. Datasets and user content (critical)
5.1 Your responsibility
You retain ownership of datasets and materials you upload (“User Content”). You grant us a limited licence to host, process, transmit, and display User Content solely to provide the Service (including to subprocessors and compute providers as needed to run jobs you launch).
You represent and warrant that:
1. You have all rights, licences, and lawful bases to upload and process the User Content for fine-tuning.
2. User Content does not unlawfully infringe third-party IP, privacy, or other rights.
3. You will not upload content that is illegal under applicable law.
5.2 Prohibited data (personal & special categories)
You must not upload datasets that contain:
| Prohibited | Examples (non-exhaustive) |
|---|---|
| Special categories of personal data under GDPR Art. 9 | Health, biometric identifiers for identification, genetic data, racial/ethnic origin, political opinions, religious/philosophical beliefs, trade-union membership, sex life or sexual orientation |
| Criminal-related personal data (Art. 10 GDPR and equivalents) | Convictions, offences, investigations |
| Personal data of third parties without a valid legal basis and required notices/consents | Customer lists, employee records, scraped private profiles, phone books of individuals |
| Children’s data where restricted by law | Data of minors under applicable age thresholds |
| Secrets / credentials | API keys, passwords, private keys, full payment card numbers, government ID numbers used as identity documents |
| Malware or abusive content | Exploit code intended for harm, CSAM, terrorist content, etc. |
Allowed by default (examples): synthetic instruction/output pairs; public-domain or licensed text you have rights to use; your own non-sensitive content; open datasets that are already lawfully redistributable and do not include prohibited categories above.
If your use case requires processing personal data of third parties or special-category data, do not use OpenReef unless you have a written enterprise agreement with us that expressly covers that processing. The self-serve Service is not designed or offered for that purpose.
5.3 Our role regarding datasets
- We process User Content as a processor on your documented instructions when you act as controller for that content (see Privacy Policy).
- We do not review every dataset manually. Detection is best-effort (technical limits, abuse reports). Violation may lead to immediate deletion of content, job cancellation, account suspension, and cooperation with authorities where required.
- You agree to indemnify us against claims arising from User Content you upload, to the extent permitted by law.
5.4 Hugging Face and third-party models
Using gated or third-party models requires compliance with Hugging Face and model licence terms. If you store an HF token, you do so under your own Hub account; you must not share tokens that grant broader rights than needed. We do not guarantee availability of any third-party model.
6. Decentralised compute (OpenGPU Network) — sub-processors
When you run a job in real / network mode (OpenGPU Network):
1. Your job configuration, dataset (or access to it), and related job metadata may be processed on third-party provider hardware operated by independent node operators (“Providers”).
2. Providers act as sub-processors (or independent processors under network rules) for the compute step. They are not our employees.
3. Providers may be located inside or outside the European Economic Area. By launching a network job you instruct us to transfer necessary data to Providers for that job, including potential international transfers. We will use appropriate safeguards where required (e.g. SCCs / network contractual terms) as described in the Privacy Policy; you acknowledge that exact physical location of a given Provider node may not be fully transparent before routing.
4. Local mode keeps training on infrastructure you control (e.g. your machine or a VPS you operate); network mode does not.
5. We are not liable for Provider downtime, hardware failure, or Provider misconduct beyond remedies available under OpenGPU / our settlement rules (credits refunds where applicable).
If you cannot accept sub-processing on third-party GPUs or possible non-EEA locations, use only local mode or do not use the Service for that data.
7. Acceptable use
You will not:
- abuse the Service (scraping, overload, reverse engineering beyond legal rights);
- attempt to bypass billing, quotas, or security;
- use outputs to violate law;
- resell the Service without permission;
- interfere with other users or Providers.
8. Intellectual property
- The Service software, branding, and documentation are ours or our licensors’.
- Adapters / trained artefacts produced from your jobs: as between you and us, you own the resulting adapter weights you download, subject to base-model licences and third-party terms. You are responsible for compliance when redistributing models.
- Feedback you give us may be used to improve the Service without obligation to you.
9. Disclaimers
THE SERVICE IS PROVIDED “AS IS” AND “AS AVAILABLE”. TO THE MAXIMUM EXTENT PERMITTED BY LAW, WE DISCLAIM WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT. We do not warrant uninterrupted or error-free operation, or that training will achieve any accuracy target.
10. Limitation of liability
TO THE MAXIMUM EXTENT PERMITTED BY LAW, OUR AGGREGATE LIABILITY ARISING OUT OF THE SERVICE IS LIMITED TO THE GREATER OF (A) AMOUNTS YOU PAID US FOR CREDITS IN THE 3 MONTHS BEFORE THE CLAIM, OR (B) EUR 50. WE ARE NOT LIABLE FOR INDIRECT, INCIDENTAL, SPECIAL, CONSEQUENTIAL, OR LOST-PROFITS DAMAGES.
Nothing excludes liability that cannot be excluded under mandatory law (e.g. intent, gross negligence, death/personal injury where applicable).
11. Indemnity
You will defend and indemnify us against claims arising from: (i) User Content; (ii) your breach of these Terms or law; (iii) your use of models/outputs in violation of third-party licences—except to the extent caused by our wilful misconduct.
12. Suspension and termination
We may suspend or terminate access for breach, risk, non-payment, or legal requirement. You may stop using the Service at any time. Provisions that by nature should survive (IP, liability, indemnity, dataset warranties) survive termination.
13. Changes
We may update these Terms by posting a new version with an updated date. Material changes may be notified in-product or by email where appropriate. Continued use after the effective date constitutes acceptance.
14. Governing law
Unless mandatory consumer law provides otherwise, these Terms are governed by the laws of Spain, and courts of [city to confirm, e.g. Madrid / Barcelona] shall have jurisdiction, without prejudice to mandatory rights in your country of residence.
15. Contact
Questions about these Terms: via the channels listed on https://openreef.network (or Telegram community link published there).