Securing

Effective Date: March 25, 2026

Web Check Us Anytime
Virus Check Us Anytime

Lisaiceland Trust Center
Welcome to the Lisaiceland Trust Center.
Active Development
Docs | ...more

Lisaiceland Fair Use & Acceptable Use Policy (FUP/AUP)

Applies to: all Lisaiceland products and MCP-powered services (including AI Voice+)
Framework: Model Context Protocol (MCP)
Last updated: March 25, 2026

1. One-Page Summary
Core Principle

Lisaiceland provides scalable AI infrastructure using MCP. Fair use ensures reliable service performance for all tenants.

Allowed

✔ customer service AI agents
✔ voice assistants and IVR automation
✔ workflow orchestration agents
✔ enterprise data chat
✔ analytics copilots
✔ voice cloning with consent
✔ multi-agent collaboration

Not Allowed

✘ abusive traffic or excessive API calls
✘ impersonation or deepfake fraud
✘ generating spam or deceptive content
✘ attempts to bypass billing or rate limits
✘ running unrelated compute workloads
✘ reverse engineering platform internals
✘ security probing without authorization

Enforcement

Lisaiceland may:
• apply rate limits
• temporarily throttle workloads
• require plan upgrades
• suspend abusive workloads
• terminate accounts in severe cases

2. Combined Acceptable Use + Fair Use Policy
2.1 Purpose

This policy ensures the Lisaiceland MCP infrastructure operates reliably, securely, and efficiently across all customers.

The policy applies to:

• AI Voice agents
• text-to-speech and speech-to-text pipelines
• conversational agents
• data retrieval systems
• API endpoints
• workflow automation agents
• analytics copilots
• knowledge agents
• multi-agent environments

2.2 Acceptable Use Requirements

Customers must:

Use services in compliance with applicable laws
Respect consent requirements for voice cloning
Avoid impersonation or deceptive identity simulation
Implement reasonable safeguards against abuse
Avoid generating harmful or misleading content
Protect credentials and API keys
Prevent unauthorized access to MCP endpoints
2.3 Prohibited Activities
Abuse of AI Systems

Prohibited behaviors include:

• deepfake impersonation without consent
• automated spam generation
• disinformation campaigns
• harassment automation
• voice cloning of individuals without permission
• social engineering or fraud automation

Infrastructure Abuse

Users may not:

• overload shared inference infrastructure
• create infinite agent loops
• generate artificial workload spikes
• mine cryptocurrency
• run general compute workloads unrelated to product use cases
• attempt to bypass usage metering
• conduct stress testing without approval

Security Violations

Customers may not:

• probe vulnerabilities
• reverse engineer MCP components
• attempt unauthorized data extraction
• exploit system latency behavior
• bypass authentication controls
• create synthetic identities to bypass limits

3. MCP-Specific Technical Usage Limits

These baseline limits apply unless overridden by contract.

Category Baseline Fair Use Range
API requests 10 – 120 requests/sec per tenant
concurrent agents 25 – 250 active agents
concurrent voice sessions 10 – 150 live calls
workflow executions 5k – 250k per day
vector database queries 50 – 500 QPS
knowledge retrieval latency target < 800 ms
audio synthesis duration 100 – 50,000 minutes/month
storage usage 10 GB – 5 TB
file uploads 1 MB – 250 MB per file
webhook delivery attempts 3 retries
orchestration depth max 10 chained agents
message context window 8k – 1M tokens (plan dependent)

Lisaiceland reserves the right to dynamically adjust limits to preserve platform reliability.

4. Rate Limit Matrix
Endpoint Type Burst Limit Sustained Limit
MCP context read 300 req/min 120 req/min
MCP context write 150 req/min 60 req/min
agent execution start 60 req/min 30 req/min
voice synthesis 30 req/min 15 req/min
speech recognition 50 req/min 20 req/min
vector search 500 req/min 200 req/min
workflow trigger 100 req/min 40 req/min
webhook ingestion 200 req/min 80 req/min
file upload 20 req/min 10 req/min

Burst limits allow temporary spikes.

Sustained limits apply over rolling windows.

5. Developer-Friendly Version
Design Best Practices
Efficient Agent Design

Recommended:

• avoid recursive loops without exit conditions
• cache repeated knowledge queries
• batch requests where possible
• minimize repeated context hydration
• avoid unnecessary agent fan-out patterns

Voice Workload Optimization

Recommended:

• pre-generate frequently used prompts
• compress audio where possible
• avoid unnecessary re-synthesis loops
• implement silence detection logic

MCP Optimization Patterns

Efficient:

agent → retrieve context → respond → terminate

inefficient:

agent → retrieve context → spawn agents → loop → rehydrate context repeatedly

Safe Multi-Agent Patterns

recommended max depth:

primary agent
→ tool agent
→ knowledge agent

avoid:

unbounded recursive orchestration

6. Monitoring & Enforcement

Lisaiceland monitors:

• token throughput
• orchestration depth
• concurrent execution load
• abnormal latency spikes
• failure rate anomalies
• suspicious usage bursts
• unusual geographic activity patterns
• high-risk voice cloning activity

Automated safeguards may include:

• adaptive throttling
• request queuing
• concurrency reduction
• anomaly detection triggers

7. Enterprise Scaling Model

Enterprise customers may request:

• dedicated inference clusters
• private MCP environments
• isolated vector databases
• custom rate limits
• reserved GPU capacity
• regional data hosting
• private voice synthesis pipelines

Custom SLAs may include:

• latency guarantees
• throughput guarantees
• uptime targets
• failover redundancy

8. SOC2-Style Control Mapping
Control Area Implementation
availability adaptive rate limiting protects uptime
security API authentication and request signing
confidentiality tenant-isolated MCP context stores
processing integrity deterministic workflow execution
monitoring anomaly detection on workload behavior
change management versioned MCP schemas
incident response automated abuse detection triggers
risk mitigation concurrency caps
data protection encrypted storage layers
auditability structured event logging
9. Enforcement Actions

Lisaiceland may:

notify customer of abnormal usage
apply temporary throttling
isolate problematic workloads
request architecture adjustments
recommend plan upgrades
suspend abusive processes
terminate repeated violations

Where feasible, customers will be notified before enforcement actions.

For additional details, please review our policies:

✅ Trust Center

Security Contact

If you have security questions or wish to report a vulnerability, please contact:

Lisaiceland Security Team

Lisaiceland
Website: https://lisaiceland.com

Email:  Submit Ticket