RPA AND AUTOMATION GOVERNANCE

Govern your bots like you govern your agents.

RPA bots operating without governance create the same risks as ungoverned AI agents: unauthorized actions, no audit trail, no kill switch. Agentomy extends its governance layer to cover bots across any RPA platform or custom automation framework.

Self-hosted by design. Bring your own model, bring your own key -- your RPA bot data + scope tokens stays in your infrastructure. Agentomy never proxies, never hosts, never sees the data plane. sovereignty model →

142

Behavioral detection patterns

Compliance frameworks

Mapped controls

Behavioral monitor methods

GovernanceBench scenarios

142 behavioral detection patterns validated against documented incidents

Every detection pattern is mapped to a real incident, a real regulatory reference, and a real detection method. The behavioral monitor includes 9 core methods available in every deployment and 31 vertical-specific methods with fleet infrastructure. The full 40-method monitor runs continuously across the bot lifecycle. No theoretical threats. No generic compliance language. Below are 10 representative patterns from the full 142-pattern library.

Runaway Loop

Runaway Bot Loop

Bot executing 100+ identical actions in 60 seconds. Indicates runaway process, infinite loop, or misconfigured retry logic.

Critical

Unauthorized Access

Unauthorized System Access

Bot attempting to access a system or endpoint not on its approved whitelist. Indicates scope creep, misconfiguration, or compromised bot.

Critical

Credential Reuse

Multiple bot instances using the same credential or service account. Breaks auditability and enables unauthorized action attribution.

High

Process Escalation

Bot attempting to trigger a higher-privilege process or escalate its own authorization tier without explicit approval.

Critical

Data Exfiltration

Configuration Drift

Bot's operational parameters changing between scheduled runs without an authorized change record in the governance trail.

High

Schedule Violation

Bot executing outside approved windows: maintenance periods, off-hours, or embargo periods. Indicates misconfiguration or unauthorized activation.

High

Configuration Drift

Data Exfiltration

Bot sending data to unapproved external destinations. Detected via destination whitelist with hash verification on every outbound call.

Critical

Lateral Movement

Bot traversing from an authorized system to an unauthorized adjacent system. Cross-system access pattern analysis detects the boundary violation.

Critical

Volume Anomaly

Unattended Spawning

Bot creating child processes or spawning new bot instances without operator awareness. Process tree monitoring catches undeclared instances.

High

Unattended Spawning

Audit Gap

Bot operating without producing evidence records. Gaps in the audit trail mean gaps in regulatory defensibility.

High

5 governance layers for the bot lifecycle

Each layer enforces one aspect of RPA governance -- from individual action validation to fleet-wide emergency halt.

Action Gate

Every bot action validated against approved process whitelist before execution. Blocks unauthorized processes, enforces frequency limits, and flags volume anomalies.

Identity Resolver

Maps every bot instance to a governance identity with unique credentials, platform origin, and permission scope. Detects shared account abuse across instances.

Schedule Guard

Enforces approved execution windows per bot. Flags and blocks out-of-hours execution. Configurable drift tolerance for legitimate operational variance.

Boundary Control

Prevents unauthorized lateral movement between systems. Every cross-system boundary crossing requires explicit authorization. Tracks traversal count per session.

Fleet Halt

One command halts all governed bots across every platform. Scoped halt by platform, process type, or risk tier. Persists across restart. Sub-20ms response time.

7 frameworks, 34 controls, real regulatory references

Every control mapping references the actual regulatory document. No generic compliance language. All mappings are self-assessed, pending external validation.

Framework	Controls	Scope
OCC Bulletin 2023-17	4	Third-party RPA vendor risk assessment, due diligence, ongoing monitoring
OCC 2011-12 / SR 11-7	4	Model risk management for AI-enhanced bots. Pure rule-based bots carved out.
FFIEC IT Handbook	3	Operations monitoring, change management, incident identification
DORA Article 6	6	ICT risk management: annual review, board accountability, 4-hour incident reporting
EU AI Act Article 6	5	High-risk classification for AI/ML bots in Annex III domains
SOX 404 + COSO RPA Framework	6	Bot lifecycle ITGCs: development, deployment, monitoring, decommissioning
PCI DSS v4.0	6	Bot identity, least privilege, audit logging, tamper detection, incident response

Real incidents, real consequences

Knight Capital Group, August 2012

$440M loss in 45 minutes

Runaway automated trading. No process execution validation, no frequency anomaly detection, no kill switch. SEC Administrative Proceeding File No. 3-15570. Agentomy detects this via the Runaway Loop pattern with auto-quarantine at 100 actions in 60 seconds.

GSA OIG Audit, August 2024

Decommissioned bots retained PII access

Seven recommendations issued after audit found decommissioned RPA bots still had access to systems containing personally identifiable information and procurement data. Report A230020/B/T/F24004. Agentomy detects this via the Unauthorized Access and Credential Reuse patterns.

Citibank, August 2020

$900M erroneous transfer

Automated process transferred the full principal instead of an interest payment. No cross-system boundary check, no authorization validation on transfer amounts. Agentomy detects this via the Process Escalation and Data Exfiltration patterns.

COSO RPA Framework, December 2024

14 control requirements published

COSO published a dedicated RPA governance framework identifying 4 governance areas and 14 specific control requirements for bots touching financial reporting. SOX 404 compliance now requires RPA-specific ITGCs. Agentomy maps all 6 SOX/COSO controls.

node-ipc Supply Chain Attack, May 2026

822K weekly downloads, credential stealer via expired domain takeover

Attacker re-registered an npm maintainer's expired email domain, hijacked their account, and published three poisoned versions. Payload fires on require() with no postinstall script -- invisible to lifecycle-hook scanners. 90+ credential categories stolen via DNS tunneling disguised as Azure infrastructure. First documented npm attack targeting AI tool configs (.claude.json). Agentomy detects this via IOC package flagging, version pinning enforcement, C2 domain detection, and obfuscated script scanning.

Three commands to governed RPA

# Install the governance adapter $ npm install @agentomy/governance # Authorize an RPA bot action $ curl -X POST http://your-agentomy-instance:3000/api/claw/authorize \ -H "Content-Type: application/json" \ -H "X-API-Key: YOUR_API_KEY" \ -d '{"agentId": "rpa-bot-invoice-processor", "action": "write", "scope": "accounts_payable", "metadata": {"botPlatform": "rpa-engine", "processName": "InvoiceReconciliation"}}' # Halt the entire bot fleet $ curl -X POST http://your-agentomy-instance:3000/api/claw/halt \ -H "Content-Type: application/json" \ -H "X-API-Key: YOUR_API_KEY" \ -d '{"reason": "runaway process detected", "operatorId": "admin-01"}'