Document status: beta draft, revised 2026-11-05. The team has not yet supplied the legal operator's registered name, correspondence address, country or data-controller identity. These documents describe current operations but are not a completed legal disclosure or a claim of commercial readiness. Contact support@unbound-me.com for clarification before purchasing or submitting personal data. Statutory rights remain unaffected.
Permitted work
Use unbound.ai for lawful development, research, education and authorized cybersecurity tasks. Security testing requires permission from the system owner and must remain within its agreed scope. An abliterated model is a model variant, not permission to misuse other people's systems or data.
Prohibited activity
- Unauthorized access, credential theft, phishing, fraudulent impersonation, extortion, destructive attacks or deployment of malware against others.
- Child sexual abuse or exploitation, non-consensual sexual material, trafficking, or instructions facilitating serious violence or other unlawful harm.
- Doxxing, stalking, unlawful surveillance, theft of personal or confidential information, or processing data without the necessary rights and legal basis.
- Spam, payment fraud, use of stolen funds, attempts to evade sanctions or other applicable restrictions, or misrepresentation of a payment's confirmation state.
- Bypassing authentication, spending protection, rate limits, custody locks or another account's permissions. Do not exploit an accidental credit or settlement error; report it.
- Overloading the service, publishing another user's keys or Seal links, manipulating referral attribution, or using support channels to distribute secrets or harmful content.
Agents and shared access
Authorize only the MCP scopes an agent needs. Review costs and requested mutations before allowing an agent to create keys or invoices. Access to a document, read scope or an API example does not authorize spending or withdrawals. You are responsible for the permissions you grant and should revoke an agent you no longer trust.
Handling reports and enforcement
Report abuse to support with a concise description and relevant request or invoice IDs. Do not send live credentials or unnecessary copies of unlawful content. We may preserve relevant evidence, restrict affected credentials, investigate, and respond to lawful requests. Actions should be proportionate to the risk; notice may be delayed where it would compromise safety or an investigation. You may request human review of a restriction through support.
Research on unbound.ai itself requires prior written scope agreement. The security reporting policy explains how to disclose an accidentally discovered vulnerability without expanding access or harming users.
