Enterprise AI Security Architecture — Part III | AetherStaff
AAetherStaff
Enterprise Agent Engineering · Security Reference
Chapter 05C · Security Architecture · Operations

Security operations: detect, contain, and recover

Part III defines the operating controls that keep enterprise AI secure after launch: network boundaries, telemetry, audit, incident response, red-team testing, governance, and recovery.

Chapter 5C defines the operational security system around enterprise AI: network controls, monitoring, audit, detection, incident response, red-team testing, recovery, governance, and the decisions required to keep security effective after deployment.

5C.1 Network and infrastructure security

Enterprise AI workloads should use explicit ingress, private service access where appropriate, controlled outbound routing, workload identity, encrypted storage, and segmented execution environments. Agents should not inherit unrestricted enterprise or internet access.

Network security zonesControlled ingress, private services, restricted egress
User / Application Zone
Gateway Zone
Agent Runtime Zone
Model / AI Service Zone
Knowledge Zone
Capability Gateway
Connector Zone
Enterprise Systems Zone
Central security telemetry · secrets · policy · certificate and key management
  • Use separate environments and identities for development, test, and production.
  • Restrict egress by destination, protocol, tenant, and workload.
  • Use private endpoints for sensitive data and model services where available.
  • Isolate high-risk code execution and file processing in disposable sandboxes.
  • Protect orchestration and policy stores with high availability and strict administration.

5C.2 Security monitoring and detection

Security monitoring should correlate application, model, retrieval, policy, tool, identity, and network signals. Traditional indicators such as authentication failure remain important, but agentic systems add suspicious loops, repeated denied capabilities, abnormal context growth, cross-domain retrieval, and unusual outbound destinations.

Detection signalPossible condition
Repeated attempts to invoke denied toolsPrompt injection, agent misconfiguration, or malicious user
Sudden increase in context volumeData harvesting, tool loop, or retrieval manipulation
Cross-tenant cache or memory accessIsolation failure
New external destination requestedExfiltration or unreviewed integration
Approval followed by different action hashTime-of-check/time-of-use violation
High rate of post-condition failuresConnector drift, target change, or model degradation
Model routing outside expected regionPolicy or gateway failure

5C.3 Audit evidence

Audit evidence should answer who initiated the task, which agent and version acted, what context and sources were used, which model processed the task, which policy decision applied, who approved the operation, what capability executed, and whether the intended post-condition was verified.

High-impact action audit record
{
  "trace_id":"tr_01J8...",
  "initiator":"user_19482",
  "agent":{"id":"payment-review-agent","version":"3.1"},
  "sources":["invoice:INV-8821","contract:CT-991:v4"],
  "model_deployment":"reasoning-enterprise-eu",
  "policy":{"id":"payment-policy-v7","decision":"allow_with_approval"},
  "approval":{"id":"ap_2201","approver":"user_7721","action_hash":"sha256:..."},
  "capability":"erp.payment.prepare",
  "operation_id":"op_44192",
  "post_condition":"payment_draft_verified",
  "outcome":"success"
}

Audit stores require integrity controls, restricted access, retention rules, and privacy minimization. Full prompt content may be necessary for some investigations, but should not be exposed broadly or retained without purpose.

5C.4 Incident response

AI incident response extends conventional security response. The organization must be able to suspend an agent, revoke its credentials, disable one capability, block one model deployment, remove poisoned knowledge, stop active workflows, and preserve evidence independently.

IncidentImmediate containmentRecovery
Prompt injection with tool attemptsDisable affected capability or agent route; preserve tracePatch content controls, policy, evaluation, and source trust
Cross-tenant disclosureStop service, revoke tokens, isolate storesCorrect partitioning, rotate secrets, notify according to obligations
Compromised connectorRevoke workload identity and network accessRebuild, verify target changes, reconcile operations
Poisoned knowledge sourceSuspend indexing and retrieval from sourceRestore trusted version and reprocess affected outputs
Unsafe model updateRoll back route or deploymentExpand evaluation and release controls
Unauthorized business actionStop workflows, reverse or compensate actionInvestigate identity, policy, approval, and connector chain

5C.5 Security testing and red teaming

Security testing should cover normal application vulnerabilities and agent-specific abuse. Test direct and indirect injection, data extraction, tool escalation, identifier substitution, cross-tenant access, approval bypass, replay, duplicate writes, malicious files, external URLs, model fallback, and recovery.

PRE-RELEASE
Controlled security evaluation

Threat-model review, SAST, dependency scanning, policy tests, connector simulation, adversarial cases, red-team exercises.

PRODUCTION
Continuous assurance

Canary monitoring, detection rules, sampled review, drift, denied tool telemetry, incident feedback, periodic access review.

Red-team findings should become regression tests. A successful attack that is only documented, but not converted into a repeatable evaluation and control, will likely recur.

5C.6 Security governance

Security approval should be risk-tiered. Read-only assistants, draft-producing workflows, and autonomous high-impact agents require different gates. Governance should cover owners, threat model, data classification, identity, network path, tool inventory, evaluation, incident owner, and retirement plan.

Every production agent has a current threat model.
Security review covers agent, model, tools, data, network, and operating process.
High-impact capabilities require separate approval and stronger monitoring.
Access reviews include agent and connector identities.
Emergency suspension and revocation procedures are tested.
Security exceptions have owners, expiry dates, and compensating controls.

5C.7 Security Architecture Decision Records

Security ADR example
# ADR-SEC-009: Prohibit unrestricted outbound HTTP tools

Status: Accepted

Context:
Agents need access to selected public APIs and approved partner services.
A generic HTTP tool would permit SSRF, data exfiltration, and unreviewed destinations.

Decision:
Agents may use only destination-specific connectors registered in the capability catalog.
All outbound traffic passes through an authenticated proxy and destination allowlist.

Controls:
- Per-destination DNS and certificate validation
- Request and response schema validation
- DLP inspection
- Tenant and agent attribution
- Rate and data-volume limits
- Security telemetry and emergency block

Consequence:
New destinations require connector onboarding, but egress becomes governable and auditable.

5C.8 Production security readiness checklist

Network zones, ingress, private access, and egress paths are documented.
High-risk execution occurs in isolated and disposable environments.
Security telemetry correlates identity, model, retrieval, policy, and tools.
Detection rules cover denied tools, exfiltration, isolation failures, and loops.
Audit evidence is tamper-resistant, access-controlled, and retention-managed.
Incident response can disable agents, models, capabilities, and connectors independently.
Poisoned knowledge can be removed and affected outputs traced.
Red-team findings become automated regression tests.
Access and security reviews include non-human agent identities.
Recovery and compensation are tested for unauthorized or partial actions.
Security exceptions expire and have accountable owners.
A full production security review is required before expanding autonomy.

5C.9 Summary

Chapter 5C turns security architecture into an operating capability. Network controls restrict reach. Monitoring and audit expose behavior. Incident response contains failures. Red-team testing converts unknown weaknesses into repeatable controls. Governance keeps the architecture effective as agents, models, and tools change.

Core conclusion: enterprise AI security is complete only when the organization can detect misuse, contain it quickly, recover business state, and prove what happened.