What ASCEND Does — And Doesn't Do

Understanding exactly what ASCEND provides — and what remains your responsibility — is critical for enterprise deployments. This document defines clear boundaries for security teams, compliance officers, and architects.

What ASCEND Does

ASCEND is a governance layer that sits between your AI agents and the systems they interact with.

Core Capabilities

Capability	Description	Implementation
Action Evaluation	Every agent action is evaluated against risk scoring and policies	`services/enterprise_risk_calculator_v2.py`
Risk Scoring	Hybrid multi-factor scoring (0-100) with CVSS/NIST/MITRE mapping	56 action types, 67 services
Policy Enforcement	Configurable rules with condition engine	`services/condition_engine.py`
Approval Workflows	Human-in-the-loop for high-risk actions	Multi-stage, SLA monitoring
Audit Logging	Immutable, hash-chained compliance logs	WORM storage, 7-year retention
Multi-Tenant Isolation	Complete data separation per organization	Row-Level Security, `organization_id` filtering

Security Controls Provided

┌─────────────────────────────────────────────────────────────────────┐
│                    ASCEND SECURITY CONTROLS                          │
├─────────────────────────────────────────────────────────────────────┤
│                                                                      │
│  ✓ Authentication (API Key + JWT + MFA)                             │
│  ✓ Authorization (6-level RBAC, 23 permissions)                     │
│  ✓ Encryption at Rest (AES-256, BYOK supported)                     │
│  ✓ Encryption in Transit (TLS 1.3)                                  │
│  ✓ Audit Trail (Immutable, hash-chained)                            │
│  ✓ Anomaly Detection (Behavioral analysis)                          │
│  ✓ Rate Limiting (Per-agent, per-organization)                      │
│  ✓ Circuit Breaker (Automatic failsafe)                             │
│                                                                      │
└─────────────────────────────────────────────────────────────────────┘

What ASCEND Does NOT Do

ASCEND Does NOT Train Models on Your Data

┌─────────────────────────────────────────────────────────────────────┐
│  ❌ ASCEND NEVER:                                                    │
│                                                                      │
│     • Uses your data to train AI models                             │
│     • Shares data between organizations                             │
│     • Stores data for purposes beyond governance                    │
│     • Accesses your data without explicit action submission         │
│                                                                      │
│  ✓ YOUR DATA IS:                                                     │
│                                                                      │
│     • Processed only for governance decisions                       │
│     • Isolated by organization (multi-tenant RLS)                   │
│     • Encrypted at rest and in transit                              │
│     • Subject to your retention policies                            │
│                                                                      │
└─────────────────────────────────────────────────────────────────────┘

ASCEND Does NOT Store Customer Secrets in Plaintext

All sensitive data is protected:

Secret Type	Storage Method	Access Control
API Keys	SHA-256 hashed + salted	Never retrievable
Database Credentials	AWS Secrets Manager	IAM role-based
Encryption Keys	AWS KMS (or BYOK)	Customer-controlled
Session Tokens	HttpOnly cookies	CSRF protected

Source: services/token_service.py, auth_utils.py

ASCEND Does NOT Replace IAM / SIEM / API Gateways

ASCEND complements your existing security stack:

┌─────────────────────────────────────────────────────────────────────┐
│                    ASCEND IN YOUR SECURITY STACK                     │
├─────────────────────────────────────────────────────────────────────┤
│                                                                      │
│   YOUR IAM (Okta, Azure AD, AWS IAM)                                │
│        │                                                             │
│        ▼                                                             │
│   ┌─────────────────────────────────────────────────────────────┐   │
│   │                        ASCEND                                │   │
│   │           AI Agent Governance Layer                          │   │
│   │   (Risk scoring, policy enforcement, approval workflows)     │   │
│   └─────────────────────────────────────────────────────────────┘   │
│        │                                                             │
│        ▼                                                             │
│   YOUR API GATEWAY (Kong, AWS API GW, Envoy)                        │
│        │                                                             │
│        ▼                                                             │
│   YOUR SIEM (Splunk, Datadog, Sentinel)                             │
│                                                                      │
└─────────────────────────────────────────────────────────────────────┘

Component	Purpose	ASCEND Integration
IAM	User/service identity	ASCEND consumes identity via SSO/SAML
API Gateway	Traffic routing, rate limiting	ASCEND provides Lambda Authorizer, Kong plugin
SIEM	Security monitoring	ASCEND exports to Splunk CIM, Datadog
Secrets Manager	Credential storage	ASCEND retrieves (never stores) secrets

ASCEND Does NOT Require Agent Redesign

Integration is minimal and non-invasive:

# BEFORE: Your existing agent code
def process_customer_request(customer_id, action):
    result = database.execute(action)
    return result

# AFTER: Same code + 3 lines for governance
from ascend import AscendClient

client = AscendClient(api_key=os.getenv("ASCEND_API_KEY"))

def process_customer_request(customer_id, action):
    # Add governance check (3 lines)
    decision = client.evaluate_action(
        action_type="database.query",
        resource="customer_db"
    )
    if decision.denied:
        return {"error": "Action blocked by policy"}

    result = database.execute(action)
    return result

Zero architectural changes required.

Data Handling Clarity

What Data Does ASCEND Process?

Data Type	Stored?	Encrypted?	Customer Controlled?	Retention
Action metadata (type, timestamp, agent_id)	Yes	Yes (AES-256)	BYOK available	Configurable
Risk scores and decisions	Yes	Yes (AES-256)	BYOK available	7 years default
Audit logs	Yes	Yes (AES-256)	BYOK available	Compliance-based
Agent credentials	NO	N/A	Never leaves your environment	N/A
Request/response payloads	NO	N/A	Stays in your infrastructure	N/A
Customer PII/PHI	NO	N/A	Never transmitted to ASCEND	N/A

What Crosses the Network Boundary?

┌─────────────────────────────────────────────────────────────────────┐
│                                                                      │
│   WHAT CROSSES TO ASCEND:          WHAT STAYS IN YOUR ENVIRONMENT:  │
│   ─────────────────────────        ─────────────────────────────────│
│                                                                      │
│   ✓ Action type                    ✗ Request payloads               │
│   ✓ Agent identifier               ✗ Response data                  │
│   ✓ Timestamp                      ✗ Database queries               │
│   ✓ Resource identifier            ✗ API responses                  │
│   ✓ Risk context flags             ✗ Customer credentials           │
│   ✓ Environment (prod/staging)     ✗ Secrets/tokens                 │
│                                    ✗ PII/PHI data                   │
│                                    ✗ Business logic                 │
│                                                                      │
└─────────────────────────────────────────────────────────────────────┘

Example: What Gets Sent

// SENT TO ASCEND (metadata only)
{
  "agent_id": "financial-advisor-001",
  "action_type": "database.query",
  "resource": "customer_accounts",
  "environment": "production",
  "timestamp": "2026-01-21T10:30:00Z",
  "context": {
    "contains_pii": true,
    "data_classification": "confidential"
  }
}

// NOT SENT (stays in your environment)
{
  "sql_query": "SELECT * FROM accounts WHERE customer_id = 12345",
  "customer_name": "John Smith",
  "account_balance": 50000.00,
  "ssn": "***-**-****"
}

Compliance Boundaries

Shared Responsibility Model

Framework	ASCEND's Role	Your Role
SOC 2	Audit logging, access controls, encryption	Data classification, user training, incident response
HIPAA	PHI access logging, encryption at rest	BAA execution, PHI handling procedures, workforce training
PCI-DSS	Cardholder data access controls, audit trails	PCI scope definition, network segmentation, key management
GDPR	Data processing records, deletion APIs	Legal basis for processing, DPO appointment, DPIA
SOX	Segregation of duties, change management logs	Financial controls design, management assertions
NIST 800-53	AC, AU, SC controls	Full control implementation, continuous monitoring

ASCEND Compliance Coverage

┌─────────────────────────────────────────────────────────────────────┐
│                    ASCEND COMPLIANCE CONTROLS                        │
├─────────────────────────────────────────────────────────────────────┤
│                                                                      │
│  ACCESS CONTROL (AC)                                                 │
│  ├─ AC-2: Account Management .............. ✓ User lifecycle        │
│  ├─ AC-3: Access Enforcement .............. ✓ RBAC (6 levels)       │
│  ├─ AC-6: Least Privilege ................. ✓ 23 granular perms     │
│  └─ AC-17: Remote Access .................. ✓ API key + MFA         │
│                                                                      │
│  AUDIT & ACCOUNTABILITY (AU)                                         │
│  ├─ AU-2: Audit Events .................... ✓ All actions logged    │
│  ├─ AU-3: Content of Audit Records ........ ✓ Who/what/when/where   │
│  ├─ AU-9: Protection of Audit Info ........ ✓ WORM, hash-chain      │
│  └─ AU-12: Audit Generation ............... ✓ Automatic logging     │
│                                                                      │
│  SYSTEM & COMMUNICATIONS (SC)                                        │
│  ├─ SC-8: Transmission Confidentiality .... ✓ TLS 1.3              │
│  ├─ SC-12: Cryptographic Key Management ... ✓ AWS KMS / BYOK        │
│  ├─ SC-13: Cryptographic Protection ....... ✓ AES-256-GCM          │
│  └─ SC-28: Protection at Rest ............. ✓ Encrypted storage     │
│                                                                      │
└─────────────────────────────────────────────────────────────────────┘

Integration Boundaries

What ASCEND Integrates With

Integration	Direction	Data Flow
Your AI Agents	Inbound	Action metadata only
Your IAM (SSO/SAML)	Inbound	Identity tokens
Your SIEM	Outbound	Audit events (Splunk CIM)
Your Notification Systems	Outbound	Alert webhooks
ServiceNow	Bidirectional	Incident tickets

What ASCEND Does NOT Integrate With

System	Reason
Your databases	ASCEND never connects directly to your data stores
Your application code	No runtime injection or bytecode modification
Your CI/CD pipelines	Governance is runtime, not build-time
Your secrets managers	ASCEND retrieves its own secrets, not yours

Frequently Asked Questions

Does ASCEND see my API request/response payloads?

No. ASCEND only receives action metadata (type, resource, timestamp). The actual request and response data stays in your environment.

Does ASCEND store my customer data?

No. ASCEND stores only governance metadata: action types, decisions, audit logs. Customer PII/PHI is never transmitted to or stored by ASCEND.

Can ASCEND block legitimate actions?

Yes, by design. That's the purpose of governance. You control the policies that determine what gets blocked. Start with monitoring mode to tune before enforcing.

What happens if ASCEND is unavailable?

Configure fail_mode:

OPEN: Actions proceed (availability prioritized)
CLOSED: Actions blocked (security prioritized)

Recommendation: Use CLOSED for production, OPEN for development.

Does ASCEND require changes to my agent architecture?

Minimal changes. SDK integration requires 3 lines of code. Gateway integration (Lambda Authorizer, Kong, Envoy) requires zero code changes.

Summary

Category	ASCEND Does	ASCEND Does NOT
Data	Store action metadata	Store request/response payloads
Security	Provide governance layer	Replace IAM/SIEM/API Gateway
Integration	Work with existing stack	Require architectural redesign
Compliance	Generate audit trails	Handle your compliance obligations
AI Models	Evaluate actions	Train on your data

Next Steps

Deployment Options — How ASCEND fits in your infrastructure
Adoption Path — Day-1 to Day-30 implementation guide
Quick Start — Get started in 5 minutes

Last Updated: 2026-01-21

What ASCEND Does​

Core Capabilities​

Security Controls Provided​

What ASCEND Does NOT Do​

ASCEND Does NOT Train Models on Your Data​

ASCEND Does NOT Store Customer Secrets in Plaintext​

ASCEND Does NOT Replace IAM / SIEM / API Gateways​

ASCEND Does NOT Require Agent Redesign​

Data Handling Clarity​

What Data Does ASCEND Process?​

What Crosses the Network Boundary?​

Example: What Gets Sent​

Compliance Boundaries​

Shared Responsibility Model​

ASCEND Compliance Coverage​

Integration Boundaries​

What ASCEND Integrates With​

What ASCEND Does NOT Integrate With​

Frequently Asked Questions​

Does ASCEND see my API request/response payloads?​

Does ASCEND store my customer data?​

Can ASCEND block legitimate actions?​

What happens if ASCEND is unavailable?​

Does ASCEND require changes to my agent architecture?​

Summary​

Next Steps​