Usage Policy

Acceptable Use Policy

This Acceptable Use Policy ("AUP") defines the binding rules, safety guardrails, and operational boundaries governing all access to and use of MarketMind AI's multi-agent orchestration platform, API gateways, vector memory stores, and integration frameworks. Compliance with this AUP is mandatory for all Enterprise Customers, developers, and end-users.

Last Updated: August 2026

1. Prohibited AI Agent Activities & Use Cases

Customer shall not configure, deploy, or execute AI agents, tool calls, or orchestration workflows through the Platform for any of the following prohibited purposes:

Autonomous Safety & System Destruction
  • ✕Executing recursive, un-throttled autonomous agent loops ("runaway agents") that invoke APIs or database writes without execution caps or terminal conditions.
  • ✕Configuring agents to alter, delete, or drop production databases, schemas, or infrastructure components without explicit human-in-the-loop (HITL) verification.
High-Risk & Illegal Application Domains
  • ✕Deploying agents to perform social scoring, biometric categorization, or real-time remote biometric identification in violation of applicable laws (including the EU AI Act).
  • ✕Utilizing AI agents to execute fully automated, non-human-reviewed decisions in high-stakes domains, including employment termination, credit evaluation, criminal risk profiling, or healthcare diagnosis.
  • ✕Generating or disseminating manipulative, deceptive, or subliminal content designed to cause material psychological or physical harm.
Malicious Cyber Operations & Impersonation
  • ✕Generating malicious code, exploits, or automated vulnerability scanners intended to bypass authorization checks or attack third-party infrastructure.
  • ✕Executing automated social engineering, spear-phishing campaigns, or deepfake impersonations of specific individuals or corporate entities without authorization.

2. API Usage Boundaries & Security Guardrails

Customer and its developers must adhere to strict technical and operational boundaries when interacting with MarketMind AI APIs and agent proxies:

Adversarial Injections & Jailbreaking

Customer shall not attempt to engineer, test, or deploy system prompts or payloads designed to execute prompt injection attacks, bypass tenant isolation guardrails, or force LLM nodes to override system safety parameters.

Rate Limits & API Quotas

Customer must respect all rate limits, token throughput caps, and concurrency limits enforced by the LiteLLM API Gateway. Circumventing gateway throttling via multi-account spoofing, distributed proxy networks, or automated scraping is strictly prohibited.

Credential & MCP Security
  • •Agents must never be assigned administrative, root, or un-scoped service account credentials.
  • •All third-party tool integrations, including Model Context Protocol (MCP) servers and external API connectors, must run with scoped, minimum-privilege credentials managed via Google Cloud Secret Manager.

3. Data Input & Sensitive Information Limits

Unless explicitly authorized under an executed Data Processing Addendum (DPA) or Business Associate Agreement (BAA), Customer shall not submit the following restricted data classes to the Platform:

Data ClassificationProcessing StatusPermitted Conditions
Protected Health Information (PHI)Prohibited (Unless BAA Executed)Requires an executed HIPAA BAA and dedicated GCP region pinning.
Payment Card Data (PCI-DSS)ProhibitedPrimary Account Numbers (PANs) or CVVs must never be ingested or stored in vector indexes (pgvector).
Government IDs & Unencrypted PIIConditionalPermitted only in tenant-isolated memory stores via approved, encrypted API payloads.
Public & Internal Domain DataPermittedCovered under standard Zero Model Training guarantees.

4. Human-in-the-Loop (HITL) & Accountability Requirements

Named Human Ownership

Every autonomous agent deployment or automated agent workflow operating within MarketMind AI must have an assigned, accountable human administrator within Customer's organization.

Consequential Approval Gates

Human authorization is mandatory prior to an agent executing:

  • •Financial transactions or binding contractual agreements exceeding pre-approved monetary thresholds.
  • •Mass external customer communications or public regulatory disclosures.
  • •Irreversible data deletion or system permission modifications.

5. Violation Enforcement & Remediation

MarketMind AI continuously monitors platform telemetry and system logs via GCP Cloud Audit Logs and LiteLLM Gateway monitoring for anomalous usage patterns. In the event of an AUP violation, MarketMind AI reserves the right to take progressive enforcement actions:

1
Automated Throttling / Warning

Real-time rate-limiting or automated warnings sent to the Account Administrator upon detection of abnormal API burst patterns or prompt injection spikes.

2
API Key Revocation

Temporary suspension of specific agent credentials or API keys involved in high-risk or prohibited activities.

3
Account Termination

Permanent account termination and IP blocking for severe, deliberate, or uncurable violations (such as malicious cyber activities or deliberate platform exploitation).

This AUP is provided for informational purposes and may be updated by MarketMind AI at any time. Continued use of the Platform after changes constitutes acceptance of the revised policy. This document does not constitute legal advice.

Questions about this policy? Contact us at legal@marketmindai.cloud