Wide latitude. Hard red lines.

BrokenGPT is designed to avoid blanket refusals for lawful, controversial, mature, political, or unconventional discussion. That latitude does not extend to using the service to harm people, exploit children, steal, or evade the law.

Prohibited use

You may not use BrokenGPT to create, request, refine, distribute, or operationalize:

• sexual content involving minors, grooming, sexualized age ambiguity, or child exploitation;
• non-consensual intimate material, sexual deepfakes, stalking, doxxing, or targeted harassment;
• credential theft, phishing, fraud, extortion, destructive malware, or unauthorized intrusion;
• concrete assistance for violent crime, terrorism, weapons construction, or illegal drug production;
• targeted threats, encouragement of imminent self-harm, or operational assistance for suicide;
• attempts to extract private data, system secrets, other customers' content, or provider credentials;
• content or activity that violates applicable law, sanctions, court orders, or third-party rights.

Platform integrity

Do not bypass quotas, share keys publicly, automate account creation, interfere with service operation, or obscure the origin of abusive traffic. You are responsible for downstream applications and must secure your API keys.

Enforcement

We may throttle, interrupt, review, suspend, or terminate activity that creates a credible safety, legal, infrastructure, or payment risk. Where appropriate, we will preserve an appeal path and explain the policy category involved. Severe or legally reportable material may be retained and reported as required by law.

Reporting

Publish a monitored abuse address before launch, then replace this line with the real contact and response-time commitment. Include the request ID, account or key prefix, and relevant timestamp in reports.