AI Governance Survey

All controls default to the recommended secure posture. BLOCK = control applied. ALLOW = risk accepted — the behavior is permitted. Toggle each control based on your organization's risk tolerance.

Claude Desktop/Cowork

Web & desktop collaboration interface

5/5

Do you want every user in the organization to have unrestricted access to Claude Cowork without any admin-level gating?

BLOCK

Control Applied

Toggle Cowork OFF organization-wide in Admin Settings > Capabilities (or use Enterprise RBAC custom roles to limit Cowork access).

AI RMF Govern 1.1 + CSF Protect PR.AC

Do you want developers to freely use Claude Cowork with full Chrome integration and unrestricted web search?

BLOCK

Control Applied

Disable the Claude-in-Chrome bridge + set org-wide web-search restrictions / site allowlists in Admin Settings > Connectors.

AI RMF Map 1.1 + CSF Protect PR.DS

Do you want users to install or connect arbitrary MCP servers / plugins in Claude Cowork?

BLOCK

Control Applied

Deploy managed-mcp.json allowlist via MDM (Jamf/Intune) so only org-approved MCPs can be used.

AI RMF Manage 2.2 + CSF Protect PR.PS

Do you want Claude Cowork to run scheduled/background tasks or Dispatch without oversight?

BLOCK

Control Applied

Disable Dispatch toggle (Admin Settings > Capabilities) and enforce folder-scoping / mount-point restrictions.

AI RMF Measure 1.2

Do you want full visibility into Cowork activity (including prompts, MCP calls, and skills)?

BLOCK

Control Applied

Enable OpenTelemetry streaming and configure log inclusion for prompts/MCP/skills (default excludes them).

AI RMF Measure 2.1 + CSF Detect DE.AE

Claude CLI

Claude Code terminal interface

3/3

Do you want developers to bypass all permission prompts in the Claude CLI (Claude Code terminal interface)?

BLOCK

Control Applied

Set "disableBypassPermissionsMode": "disable" + "allowManagedPermissionRulesOnly": true in managed-settings.json.

AI RMF Manage 1.3

Do you want unrestricted Bash, file read/write, or network commands in the CLI?

BLOCK

Control Applied

Add explicit deny/ask rules in managed-settings.json (e.g., deny "Bash(curl *)", "Read(**/.env)", "Write(**)" with ask).

CSF Protect PR.PS + AI RMF Map 2.2

Do you want users to run custom hooks or non-managed hooks in Claude CLI?

BLOCK

Control Applied

Set "allowManagedHooksOnly": true in managed-settings.json.

AI RMF Govern 2.1

Claude Code in Web Browser

Cloud-hosted coding sandboxes

2/2

Do you want Claude Code (web) sessions to have unrestricted network access in the cloud sandbox?

BLOCK

Control Applied

Force "No network access" or custom restricted-domain environment (default allowlist is broad).

AI RMF Manage 1.1 + CSF Protect PR.IP

Do you want developers to run Claude Code web with full filesystem and credential exposure risk?

BLOCK

Control Applied

Use the isolated cloud sandbox + enforce no sensitive credentials in the session (proxy handles git auth).

AI RMF Map 1.5

Claude in Chrome

Browser extension for page interaction

3/3

Do you want the Claude Chrome extension enabled for anyone in the organization?

BLOCK

Control Applied

Toggle the extension OFF org-wide in Admin Settings > Capabilities (or use site blocklist).

AI RMF Govern 1.2 + CSF Protect PR.AC

Do you want Claude in Chrome to interact with any website without restriction?

BLOCK

Control Applied

Configure site allowlist + blocklist in admin console (Team/Enterprise only).

CSF Protect PR.DS

Do you want autonomous (hands-off) mode in Claude in Chrome?

BLOCK

Control Applied

Force default "Ask Before Acting" mode; disable pre-approve for high-risk actions (purchases, deletes).

AI RMF Manage 2.3

Codex Desktop

Computer-use desktop agent

3/3

Do you want Codex Desktop to have unrestricted computer-use (screen/keyboard/mouse control) without approvals?

BLOCK

Control Applied

Enforce approval_policy in config.toml + require macOS TCC prompts (Screen Recording + Accessibility) per-session.

AI RMF Manage 1.3 + CSF Protect PR.AC

Do you want Codex Desktop plugins or 90+ marketplace extensions installed without vetting?

BLOCK

Control Applied

Restrict to private/approved marketplace only via team config.toml.

AI RMF Map 2.1

Do you want Codex Desktop to store credentials or run with full session privileges?

BLOCK

Control Applied

Set sandbox_mode to :read-only or :workspace and protect writable roots in config.toml.

CSF Protect PR.DS

Codex CLI

Command-line coding agent

4/4

Do you want Codex CLI to auto-execute without approval (full access mode) on developer machines?

BLOCK

Control Applied

Force approval modes: read-only or auto (with explicit workspace boundaries) in config.toml layers.

AI RMF Manage 1.3

Do you want Codex CLI to trust and auto-load MCP servers from project-local config files?

BLOCK

Control Applied

Disable project config loading for untrusted repos or enforce system-level MCP allowlist only.

CSF Protect PR.PS + AI RMF Map 1.3

Do you want Codex CLI to have unrestricted network or filesystem access outside the workspace?

BLOCK

Control Applied

Set sandbox_workspace_write.* protected paths + network allowlist in config.toml.

AI RMF Manage 2.2

Do you want Codex CLI to override environment variables or CODEX_HOME from project .env files?

BLOCK

Control Applied

Lock cli_auth_credentials_store and prevent CODEX_HOME redirection via managed config.

CSF Protect PR.IP

20 controls will be enforced in your deployment.

Ready to generate your policy and receive your governance package?