This policy applies to every workspace, employee, API key, and public page. It exists so humans stay accountable for what an AI employee drafts or sends after approval.
You may
- Hire employees for your own research, drafts, support replies, and internal ops.
- Connect model keys and tools you are licensed to use.
- Publish public or unlisted employee pages that describe a job, not secrets.
- Fork a public employee into your workspace and run it on your keys.
You may not
- Break the law, or use an employee to do so (fraud, phishing, unauthorized access, malware).
- Upload or generate child sexual abuse material, or sexual content involving minors.
- Target or harass people, or produce violent or extremist operational content.
- Bypass another service’s access controls, scrape in violation of its terms, or scan private networks (SSRF is blocked for a reason).
- Spam, spoof, or send commercial email or Slack without a lawful basis and a human approval when the skill is gated.
- Attempt to extract other customers’ data, credentials, or run logs.
- Resell the service as a marketplace or white-label without a written deal.
- Overwhelm the runtime (automated run storms beyond plan limits, or probing for exploits).
- Present model output as professional legal, medical, or financial advice without a qualified human.
Approvals
Turning off every approval gate does not make prohibited use allowed. If a skill is marked dangerous, the run should pause. You must not use HTTP or webhooks to evade that pause.
Enforcement
We may reject runs, disable skills, suspend a workspace, or report illegal content. Contact [email protected] to report abuse.