Acceptable Use Policy

This policy applies to everyone using the service, including the end users of an assistant you deploy. If you operate an assistant, you are responsible for its use by your users.

• Break the law. Anything unlawful in the jurisdictions that apply to you or to us.
• Harm people. Threats, harassment, stalking, or content that sexualizes children. Content promoting self-harm or violence against others.
• Deceive at scale. Impersonating a real person or organization, generating fake reviews or records presented as genuine, or producing disinformation campaigns. Building a thing that pretends to be a person to extract money or credentials.
• Attack systems. Malware, intrusion, credential stuffing, denial-of-service, or scanning and exploiting systems you are not authorized to test.
• Evade our limits. Circumventing rate limits, quotas, authentication, or routing controls; reselling raw model access obtained through us as if it were your own inference product.
• Mishandle sensitive data. Sending categories of data your plan and configuration are not set up for — for example regulated health or payment card data — without agreeing that scope with us first.
• Make consequential decisions automatically. Using output as the sole basis for decisions with legal or similarly significant effects on someone — credit, employment, housing, insurance, or legal status — without meaningful human review.

Our published measurements may be read, quoted, and cited with attribution. Do not scrape the site systematically, republish the dataset as your own, or present our results as your own measurements. If you want the data in bulk for research, ask — the answer is usually yes.

Our first step is normally to contact you, because most problems are configuration mistakes rather than intent. Where the risk is immediate — an active attack, illegal content, or something that threatens the service for other clients — we may suspend access first and explain immediately afterwards.

Repeated or deliberate breaches can result in termination under the Terms of Service. We will not terminate a paid account for a first, remediable issue without giving you the chance to fix it.

To report misuse of the service or an assistant operated through it, email [email protected]. Include enough detail to identify the assistant or account, and we will investigate.