Version 2026-10-19 · Effective 2026-10-19

1. The rule

This policy applies to your account, your workspace, your agent and everything you run inside the sandbox. Using the service means you accept it. It forms part of the Terms of Service, which explain the order of precedence between our documents.

In short: do only lawful work you are allowed to do, keep it inside your own sandbox, and do not use the service to hurt anyone.

2. Prohibited activities

You may not:

• Use the service for criminal activity, or to plan, prepare or carry it out.
• Introduce, run or distribute malware, ransomware, or code that damages or interferes with the platform, another tenant's sandbox, or any third-party system.
• Attack, scan, probe, overload or interfere with the platform, its gateway or its network, or another tenant's workspace or data, including through the agent or your own scripts.
• Send phishing messages, or collect, steal, guess or buy credentials, tokens, keys or session cookies.
• Store, request, produce or distribute child sexual abuse material, or any content that is illegal, or that abuses, harasses or defrauds a person.
• Send bulk unsolicited messages, or run spam campaigns from the service.
• Mine cryptocurrency, or use the workspace for work the plan is not meant for.
• Scrape or crawl content in breach of the law or of a third party's terms, or collect personal data without a lawful basis.
• Circumvent or probe the usage, cost or security limits, including by running several workspaces to avoid a cap, or by telling the agent to work around its own guardrails.
• Resell, share or lend an account or a workspace, or let more than one person use one account.
• Use the service to build a product or dataset that violates a third party's rights, including copyright, database rights, or a model provider's terms.
• Use the service for high-risk professional decisions, medical, legal, financial or safety-critical, without a qualified professional involved. You may use it to prepare material for a professional, not to replace one.
• Train, fine-tune or evaluate another model on the service, or on content in it, without our written consent.
• Publish or expose your own secrets, such as a model API key, a token, a private key or a password, or write them outside your workspace.
• Store or handle content you have no right to store or handle.

3. What we may do about a breach

Depending on how serious the breach is, we may:

• ask you for information about what you are doing;
• reduce a limit, such as active hours, concurrency or a spend cap;
• suspend a workspace or an account, in whole or in part;
• terminate the account and delete the workspace;
• cooperate with a law enforcement or other authority where the law requires it, or where the platform, our customers or the public need protection.

Where we can, we tell you what we found, what we did, and what you can do to fix it. We may act first where delay would cause harm.

4. Our responsibilities

• We act reasonably and proportionately. We do not treat a mistake like a crime, and we do not treat every mistake the same way.
• We tell you what we did: which action we took, roughly when, and why, unless the law stops us or the answer would harm someone.
• We give you a chance to respond before we end an account, except where the breach is serious, where continuing would cause harm, or where the law does not allow time.
• Sensitive actions are written to the audit log, and that log is the first place we look.

5. If we suspend you for something you did not do

If we suspend or terminate your account for a breach you did not commit, you get a pro-rata refund of the unused part of the period you paid for. The same applies where the suspension was caused by a third party's action or by an error on our side. We do not charge for time your workspace was suspended for a reason that was ours.

6. Chargebacks

If you reverse a payment by chargeback without talking to us first, and we do not reverse the refund, we may apply the amount to your future use of the service, or refuse to start a paid plan until it is settled. Before we do that, we ask you once through the support channel to settle the amount.

7. Right to appeal

• Send the appeal through the in-platform support channel, or as a formal notice to {{SUPPORT_EMAIL}}.
• Say which action you are appealing and what you think happened. We look again at the job receipts, the audit log and the account history.
• We answer the appeal and say what the outcome is. Where we accept it, we undo the action where we can, and correct what the suspension caused.
• We look at appeals promptly, but an appeal does not automatically lift a suspension while we review it.

8. Monitoring

We are not required to watch every part of the service all the time. We do not read your workspace or your jobs unless you ask us to and give permission, or where the law requires it, or where we must act to stop harm.

We may inspect what runs in a sandbox, and what it connects to, to keep the platform safe and to enforce this policy.

9. Governing law and jurisdiction

This policy is governed by the law of {{GOVERNING_LAW}}, and the courts of {{COURTS}} have exclusive jurisdiction. Nothing here removes a right you have under the mandatory law of your country, including your rights as a consumer. In the Kingdom of Saudi Arabia, the Arabic version of this policy prevails over the English version.

Version history