Opencontroller by Lyzr
Opencontroller blog
Answers to the questions enterprises ask about governing AI agents
Agent governance basics
All in this topicCan one governance program cover classic ML models and AI agents?
Yes, with one policy layer and different controls underneath. A side-by-side of inventory fields, tests and runtime controls for ML models, LLM apps and agents.
What should a CISO require before an AI agent goes to production?
Twelve go-live gate criteria for AI agents, each with the evidence artefact that satisfies it, plus a three-tier risk model and post-launch triggers.
What should an AI agent registration questionnaire ask?
Fifteen registration questions for AI agents grouped by purpose, data, actions, identity and ownership, and how each answer sets the agent's risk tier.
What can AI governance do with read-only access versus being in the traffic path?
Registry, observe and enforce: what AI agent governance can do with read-only API access, with platform admin rights, and inline in the traffic path.
Should production and test AI agents be governed in one control plane?
Why one AI agent control plane should cover dev, test and production, with an environment-tagging scheme and the policies that differ per environment.
Discovery and shadow agents
All in this topicWhich AI asset types should an AI inventory cover?
The ten AI asset types an enterprise inventory should cover, beyond agents: models, keys, MCP servers, skills, GPTs, coding agents, vector stores and more.
How do you discover AI usage when you cannot get admin API keys?
No provider admin key? Six other signals that reveal AI usage and spend (cloud billing, cloud APIs, IdP logs, egress logs, secret scanning, expenses) and what each misses.
How do you discover employees' AI use in the browser?
Four ways to find browser-based AI use — SWG/CASB logs, managed browser reporting, extensions, enterprise workspace admin data — what each sees and privacy limits.
Agent identity and access
All in this topicShould AI governance tools allow local logins or SSO only?
Why AI governance tools should be SSO only, with SCIM provisioning, IdP group role mapping, a five-role model and two break-glass accounts.
What happens to AI agents when their owner leaves the company?
Orphaned AI agents keep running on a leaver's keys and access. A step-by-step leaver workflow to find, reassign and re-credential every agent they owned.
How do you assign an owner to every AI agent using your identity provider?
Map every AI agent to an owner in your IdP, apply joiner/mover/leaver rules to agents, and find orphaned agents when owners leave.
Runtime control
All in this topicHow do you test AI agents with simulated users before production?
How to build a persona and scenario matrix, set pass thresholds with repeated runs, and how simulated-user testing differs from red-teaming AI agents.
Which AI agent actions need human approval?
A scoring matrix for deciding which AI agent actions need human approval, by reversibility, value, data class and blast radius, with timeout rules.
How should you set severity levels for AI agent policies?
A severity rubric for AI agent policy findings based on data sensitivity, action reversibility and blast radius, with response times and example findings.
How do you export AI agent traces with OpenTelemetry?
Export AI agent traces with OpenTelemetry GenAI conventions: span names, key gen_ai attributes, content capture rules and their Development status.
Tools and MCP
All in this topicWhat is agent skills governance?
Agent skills governance defined: controls over the SKILL.md packages agents load, across authoring, review, versioning, distribution, revocation and usage.
How do you review MCP servers your agents already use?
A ten-point review checklist for MCP servers you have discovered: publisher, version pinning, transport, auth, tool descriptions, write tools, data reach and owner.
Gateways and cost
All in this topicHow do you see AI costs inside data platforms like Snowflake and Databricks?
The system views and billing tables in Snowflake and Databricks that expose AI functions, agent and model serving spend, and how to roll them up for finance.
How do you monitor LLM provider latency and errors across agents?
A per-provider SLO set for LLM traffic: p95 latency, time to first token, 429 and 5xx rates, plus the fallback triggers that should act on them.
How do you manage IP allow-lists for an LLM gateway without infrastructure drift?
Stop LLM gateway allow-list drift: one owner per list, managed CIDR lists fed by vendor egress feeds via pull requests, and scheduled drift checks.
What network setup does a self-hosted LLM gateway need?
A ten-step network checklist for a self-hosted LLM gateway: DNS, private CA certificates, ingress, egress allow-lists, proxy bypass and owners.
What AI spend reports does a CFO actually need?
Five AI spend report definitions for finance, with grain, fields, owner and cadence, separating seat, consumption and infrastructure spend.
Where do AI agent costs usually leak?
Eight AI agent cost leak patterns, from retry storms to orphaned API keys, with the metric that exposes each and a monthly cost-leak review.
How do you set budgets and alerts for AI agent spend?
A four-level budget hierarchy for AI agent spend (organisation, team, agent, version) with alert thresholds, hard-stop rules and what provider limits miss.
How do you test whether a cheaper model can replace the one an AI agent uses?
A replay-and-judge method to test a cheaper model for an AI agent: sample traffic, score with LLM judges, set acceptance thresholds, then canary live traffic.
Evidence and frameworks
All in this topicWhat evidence will a regulator ask for about your AI agents?
An inspection-ready evidence pack for AI agents: each artefact, its source system, refresh frequency and owner, with EU AI Act and DORA reference points.
How do you scope an AI agent red-team assessment?
A scoping template for AI agent red-team assessments: targets, rules of engagement, attack classes mapped to OWASP and MITRE ATLAS, deliverables and retest terms.
Should you write custom AI agent policies or rely on framework mappings?
When built-in OWASP, NIST and ISO 42001 mappings are enough, when to write your own AI agent policies, and five example organisation-specific rules.
How long should you keep AI agent logs and traces?
Retention periods for AI agent logs, traces, prompts and policy events, with the EU AI Act six-month minimum, PCI DSS 12 months and the Omnibus dates.
What AI agent controls do banks need now that model risk guidance excludes agentic AI?
SR 11-7 was rescinded in April 2026 and the revised guidance carves out generative and agentic AI. The controls banks should keep for AI agents anyway.
What AI governance evidence do insurers need for regulators?
Evidence US insurers should keep for the NAIC Model Bulletin on AI: written AIS Program, inventory, testing, monitoring, vendor records — extended to AI agents.
Buying and comparison
All in this topicCan you buy AI governance software through a cloud marketplace?
How AWS and Microsoft Marketplace private offers work, when a purchase counts toward your cloud commitment, and the questions procurement should ask first.
How do you deploy an AI governance platform on a private Kubernetes cluster?
Step-by-step: mirror images and charts, pin digests, set workload identity and cross-account roles, and sync an AI governance platform with Argo CD.
What should an RFP for AI agent governance include?
RFP requirement wording for AI agent governance across eight sections, with must/should levels, the proof to request and a scoring model.
Do you need a separate AI governance deployment per region?
Single, hub-and-spoke or fully regional: how to decide how many AI agent governance deployments you need, based on residency, latency, regulators and cost.
How should you run a pilot of an AI agent governance platform?
A two-to-four-week pilot plan for an AI agent governance platform, with week-by-week scope and measurable success criteria per capability.
What security questions should you ask an AI agent governance vendor about itself?
A governance tool sees your whole agent estate. 20 security questions to ask the vendor about data egress, deployment, access, logs and its own AI use.
Platform Guides
All in this topicHow do you govern Cursor across an engineering organisation?
Cursor controls by plan: SSO, Privacy Mode, run modes, MCP allowlists, hooks, MDM policies, audit logs and the Admin API, plus what they do not cover.
How do you govern Claude Code across an enterprise?
Managed settings locations and precedence, permission deny rules, sandbox, MCP and hook lockdown, and OpenTelemetry monitoring for Claude Code at scale.
How do you govern GPTs and workspace agents in ChatGPT Enterprise?
RBAC, app and connector controls, agent mode, compliance logs and the GPT retirement timeline for ChatGPT Enterprise admins, with what stays out of scope.
How do you govern AI agents built on Databricks?
Unity Catalog grants for agent tools and models, Unity Gateway policies and usage tables, MLflow judges and tracing, and the gaps for agents on Databricks.
How do you govern SAP Joule agents?
How Joule inherits SAP user roles, what Cloud Identity Services, the Agent Gateway and generative AI hub add, and where an estate-wide registry fits.
How do you govern Salesforce Agentforce agents?
Agent user permissions, Einstein Trust Layer, audit trail in Data 360, Testing Center and session tracing for Agentforce, plus what sits outside Salesforce.
How do you govern agents deployed on Vertex AI Agent Engine?
Agent Identity, IAM, VPC Service Controls, CMEK, audit logs, tracing and Model Armor for Vertex AI Agent Engine (now Agent Runtime), with the gaps.
How do you govern Microsoft Copilot Studio agents?
Environment routing, environment groups, data policies, sharing limits, Entra Agent ID, inventory and Purview audit for Copilot Studio agents, in order.
How do you govern agents in Microsoft Foundry?
Foundry Agent Service identity, networking, guardrails, evaluators, tracing and Control Plane mapped to a checklist, with what is preview and what is left out.
How do you govern AI agents built on Amazon Bedrock AgentCore?
Map AgentCore Identity, Gateway, Cedar Policy, Observability and Evaluations to a governance checklist, and see what they leave uncovered.
How do you track Claude spend across Amazon Bedrock, Claude Platform on AWS and the Anthropic API?
Where Claude costs appear for Amazon Bedrock, Claude Platform on AWS and the Anthropic API, which credential each needs, and how to attribute spend to agents.
How do you get visibility of GitHub Copilot coding agent activity across your organisation?
Where GitHub shows Copilot cloud agent activity — AI Controls, agent sessions, audit log, usage metrics API — what each misses, and how to fill the gaps.
How do you track Snowflake Cortex AI spend by agent or team?
The ACCOUNT_USAGE views that hold Snowflake Cortex spend, the columns for attribution, and SQL to roll credits up by agent and cost centre.
Put the framework into practice