razorBridge Acceptable Use Policy

Applies to: razorBridge AI Blades, events, razorBridge AI Gateway and every other razorBridge service run by KlusAI Labs SRL ("we", "us"). Contact: [email protected]

This policy is part of the razorBridge B2B Terms. It says what you may not do with our GPUs and models, and what we do about it. "You" means the business or public body that holds the account. You are responsible for everyone who uses it, including staff, students and event participants.

1. Crypto mining

You may not mine, stake, farm or validate any cryptocurrency or token on our GPUs. This includes proof-of-work, proof-of-space and "useful work" schemes, test runs, benchmarks of mining software, and mining hidden inside another workload.

We treat mining as a serious breach (section 8). We do not refund the compute it used.

2. Abuse of our service

You may not:

• open several accounts, or use several emails, to get more vouchers, more event credit or higher limits than one account gets;
• get around a limit, a stop at €0, a suspension or a residency setting;
• share your login or API keys outside your organisation, except with participants of an event you run through razorBridge;
• interfere with other customers, our systems or our providers' systems, or try to reach data that is not yours;
• run workloads designed to tie up GPUs without using them, to deny them to others.

3. Illegal and harmful content

You may not use razorBridge to create, store, train on, or distribute:

• child sexual abuse material, or any content that sexualises minors. We report this to the authorities and keep the evidence the law requires;
• terrorist content, or content that incites violence;
• content that infringes someone else's copyright, trade marks or other rights;
• malware, ransomware or other malicious code meant for use against others;
• anything else that is illegal in Romania, in the EU, or where you are established.

You may not use razorBridge for an AI practice banned by Article 5 of the EU AI Act (for example, social scoring, or untargeted scraping of facial images to build recognition databases).

4. Attacks and scanning

You may not use a blade to:

• attack any system, including denial-of-service attacks, password guessing or exploitation;
• scan networks or ports, or probe for weaknesses, on systems you do not own;
• send spam or phishing, or run open proxies, open relays, VPN exit nodes for the public, or botnet controllers;
• hide where traffic comes from in order to commit any of the above.

Security testing is allowed only against systems you own, or systems whose owner has given you written permission. Keep that permission and show it to us if we ask. Tell us first, at [email protected], if the test will generate heavy traffic.

5. Resale

You may not resell, rent out or sublicense razorBridge GPUs, models, credit, accounts or API keys to third parties without our written agreement.

Allowed without asking:

• your own staff and contractors using your account for your work;
• an organiser giving event participants access to blades as part of an event run through razorBridge;
• selling your own product or service that runs on razorBridge, as long as your customers cannot use our GPUs or models directly as a general compute service.

6. Training data and models

• You need the rights. You must have the right to use all data you train, fine-tune or run on. This includes copyright and database rights, and any licence terms or rights reservations (such as text-and-data-mining opt-outs) that apply to the data.
• Personal data. If your data includes personal data, you are the controller. You need a lawful basis and must meet your GDPR duties. Do not process special-category data (health, biometric data and the like) unless you have checked that you are allowed to. Our Data Processing Agreement covers our role.
• Model licences. You must follow the licence of any model you download, fine-tune or serve, including any limits on use.
• AI Gateway. You must also follow the usage policy of each model's provider. Models not labelled EU-only may be processed outside the EU. Do not send them data that must stay in the EU.
• Your outputs. We claim no rights in your data, models or outputs, and do not train models on your data.

7. Events

Organisers must tell participants that this policy applies and must act on a breach by a participant. We may remove a participant from an event without stopping the whole event.

8. What we do about a breach

We choose the step that fits the breach. Where we can, we warn you first and give you time to fix it.

Teardown deletes data. When a blade is torn down, its disks are deleted and cannot be recovered. Keep your own copies.

We may keep logs and evidence where the law requires it or to protect our rights, report illegal activity to the authorities, and give our hosting provider what it needs when it asks about abuse from our network.

If we suspend you by mistake, we restore access and return the charge for the lost compute time to your balance as credit.

We may close the account for a serious or repeated breach (B2B Terms, section 12). We do not refund credit used to breach this policy.

9. Reporting abuse

Write to [email protected], subject "Abuse", with:

• what happened, and when (with the time zone);
• the IP address, hostname or API key prefix involved;
• logs or other evidence;
• how we can reach you.

We acknowledge reports within 2 working days and act faster on serious cases. Law-enforcement requests go to the same address; we respond as the law requires.

If you think we acted on a report about you by mistake, reply to our notice. A person reviews every objection.

10. Changes

We give 30 days' notice of a material change by email, unless the law or a new type of abuse requires a faster change.