1. AI Philosophy & Principles
We build AI agent infrastructure with capability and safety as equal priorities. Agent autonomy must remain observable, auditable, and bounded.
2. Strict No-Training Policy
We do not train foundation models using your prompts, configurations, run logs, or payload data. API providers are used under their developer API commitments.
3. Safety & Abuse Boundaries
Prohibited uses include cyber abuse, scams, deceptive disinformation, platform bypass attempts, and uncontrolled self-replication patterns.
4. Human-in-the-Loop (HITL) Guidelines
High-risk actions (for example destructive writes, payments, or arbitrary code execution) should require human verification and explicit approval gates.
5. Tool & MCP Security
- Sandboxed execution boundaries and timeout enforcement.
- Encrypted secret injection with least-privilege recommendations.
- Runtime isolation to prevent cross-tenant access.
6. Built-in Agent Evaluations
Teams should continuously test behavior, regressions, and safety constraints before promoting agent changes to production.
7. Prompt Assets & Output Intellectual Property
You own your prompts, workflows, and generated outputs. Anter does not claim rights to your agent configurations.
8. Auditing & Enterprise Compliance
For enterprise governance needs, contact security@anter.ai.