Benchyard Acceptable Use Policy

Version: 0.2 | Last updated: October 10, 2026

This Policy is part of the Terms of Service. It applies to activity you initiate or authorize through Benchyard, including agents, terminals, scheduled jobs, connected machines, tools, and network traffic.

1. Lawful and authorized use

Use Benchyard only for lawful activity and only on systems, accounts, data, and resources you are authorized to access. Obtain required permissions before processing another person's data or intellectual property. Follow applicable sanctions, export controls, and provider eligibility rules. Benchyard access does not override a connected provider's location, ownership, account, or usage restrictions.

2. Prohibited conduct

You must not use the Service to:

• Steal credentials or data, gain unauthorized access, bypass another user's permissions, or exploit another person's infrastructure without authorization.
• Develop, deploy, or operate malware, ransomware, botnets, credential-stealing tools, denial-of-service attacks, or other capabilities for harmful or unauthorized activity.
• Send spam, run phishing or fraudulent impersonation campaigns, distribute scams, or evade abuse-prevention measures.
• Infringe intellectual-property, confidentiality, publicity, or privacy rights, or unlawfully disclose personal data, including doxxing.
• Create, distribute, or facilitate child sexual abuse material, exploitation of children, credible threats of violence, trafficking, or other illegal exploitation.
• Facilitate unlawful discrimination, persecution, prohibited surveillance, or consequential automated decisions about people without required legal safeguards and meaningful human oversight.
• Operate a safety-critical system in a manner where an agent's error could reasonably result in serious physical harm, without an expressly agreed and appropriately controlled use case.
• Escape a workspace boundary, access another customer's data, interfere with the Benchyard control plane, or disable or falsify resource metering, audit, or security controls.
• Evade limits, payment obligations, suspensions, or geographic restrictions through multiple accounts, account sharing, proxying, or misrepresentation.
• Pool or resell personal model-provider subscriptions, transfer provider credentials unlawfully, or represent access as provider-approved when it is not.
• Run unauthorized cryptocurrency mining, a public open proxy or anonymizing relay, unsolicited bulk scanning, or another workload outside the permitted purpose or disclosed resource limits.

Installing legitimate security tools, testing your own software, or conducting research with valid authorization is not prohibited merely because similar techniques can be misused. Testing Benchyard's infrastructure requires our prior written authorization unless expressly permitted by a published vulnerability-disclosure program or applicable law.

3. Sensitive data and credentials

Do not place regulated health records, payment-card authentication data, classified material, or other specially regulated data into the Service unless an appropriate written agreement and safeguards are in place. Use secrets and credentials with the minimum permissions needed. You must not place another person's credentials in an agent session without authority or invite an agent to disclose credentials to an unauthorized recipient.

4. Resource and agent controls

Remain within the machine, storage, network, concurrency, and account limits disclosed for your service. Automated activity must have a legitimate purpose. Stop schedules and background tasks you no longer need. Review permissions before allowing an agent to edit, execute, publish, deploy, or communicate externally. A feature allowing privileged execution does not authorize interference with infrastructure belonging to us or another customer.

5. Reports and enforcement

Report suspected abuse or a security issue to hello@benchyard.com. Include a resource identifier, relevant time, a description, and evidence you are entitled to share. Do not send passwords, private keys, or unnecessary personal data. Copyright reports follow the Copyright and Complaints Policy.

We may investigate credible reports and take proportionate measures, including warning you, restricting a connection or workload, removing prohibited material, or suspending or terminating access. We may preserve relevant evidence where legally justified. We limit access to content to what is reasonably necessary for the investigation and applicable obligations; we do not promise comprehensive monitoring of every task.

Where practicable, we will explain the action and provide an opportunity to correct an issue. Immediate action may be needed for ongoing harm, serious security risk, or legal requirements. You may seek review by contacting hello@benchyard.com. We will assess the relevant facts and applicable permissions, including authorization for legitimate security research.