ACL Tiers¶
Access Control Level (ACL) tiers determine the strictness of governance applied to an agent.
Overview¶
| ACL Tier | ARS (Agent Risk Score) | Oversight Level | Typical Use Case | Latency |
|---|---|---|---|---|
| ACL-0 | 0-2 | Minimal | Read-only bots, info retrieval | ~10ms typical, <50ms max |
| ACL-1 | 3-4 | Light | Scripted agents, simple tasks | ~20ms typical, <100ms max |
| ACL-2 | 5-7 | Standard | Customer service, moderate autonomy | ~50ms typical, <150ms max |
| ACL-3 | 8-10 | Enhanced | Business processes, high autonomy | ~100ms typical, <200ms max |
| ACL-4 | 11-13 | Strict | Financial systems, critical ops | ~200ms typical, <350ms max |
| ACL-5 | 14-15 | Maximum | Mission-critical, life-safety | ~500ms typical, <1000ms max |
Tier Details¶
ACL-0: Minimal Oversight¶
When to use: Read-only agents, information retrieval, low-risk actions
Characteristics: - Basic trace validation - Minimal latency (~10ms typical, <50ms maximum) - Simple logging - No complex evaluation
Example: FAQ bot, document search assistant
ACL-1: Light Oversight¶
When to use: Scripted agents with predictable behavior
Characteristics: - Pattern matching evaluation - Quick validation (~20ms typical, <100ms maximum) - Basic intervention logic - Standard logging
Example: Appointment scheduler, data entry assistant
ACL-2: Standard Oversight¶
When to use: Production agents with moderate autonomy (RECOMMENDED for most use cases)
Characteristics: - Full cognitive trace evaluation - Proportionate interventions - Comprehensive logging - Trust score tracking
Example: Customer service agent, sales assistant
ACL-3: Enhanced Oversight¶
When to use: High-autonomy agents handling important business processes
Characteristics: - Deep reasoning analysis - Multi-factor evaluation - Advanced intervention logic - Detailed audit trails
Example: Business process automation, data analysis agents
ACL-4: Strict Oversight¶
When to use: Critical systems with significant consequences
Characteristics: - Comprehensive evaluation - Human-in-the-loop for escalations - Cryptographic audit logs - Real-time monitoring
Example: Financial trading agent, healthcare assistant
ACL-5: Maximum Oversight¶
When to use: Mission-critical and life-safety systems
Characteristics: - Maximum evaluation depth - Mandatory human approval for all actions - Distributed consensus validation - Immutable audit logs - Real-time compliance checking
Example: Autonomous medical systems, critical infrastructure
Choosing Your ACL Tier¶
Start with ACL-2
Most production agents should start with ACL-2 (Standard Oversight). It provides good balance between governance and performance.
Interactive Wizard¶
Answer three quick questions to get a personalized recommendation:
What can your agent do?
ARS Calculator¶
For a more detailed assessment, use the ARS Calculator to determine the appropriate tier for your agent.
Dynamic Re-Tiering¶
Agents can automatically move between ACL tiers based on their trust debt:
- Trust debt increases → Higher ACL tier (more oversight)
- Trust debt decreases → Lower ACL tier (less oversight)
See Trust System for details.