AI Settings Page — Scheduled Penetration Tests and Category Allocation

How to use the redesigned AI settings page to schedule autonomous penetration tests, manage category allocation, and configure AI spend as separate, role-appropriate controls.

AI Settings Page — Scheduled Penetration Tests and Category Allocation

The AI settings page organises autonomous-penetration-test scheduling, AI spend allocation, and budget management into three distinct cards. Each card has its own edit modal and its own permission requirements, so operators can access only the controls they are entitled to configure without navigating a shared, super-admin-gated wizard.


Cards on the AI settings page

Schedule Autonomous Penetration Test

This card controls whether Guard automatically initiates hunts on a recurring schedule and which hunt profile is used.

Who can edit: Users with the manage-settings role.

Configurable options:

Option

Description

Automatic scheduling toggle

Enable or disable recurring autonomous penetration tests.

Hunt profile

Select the profile that determines scope and behaviour for scheduled runs.

Deep-link support: You can link directly to this card with a URL parameter. This is useful for embedding links in runbooks or notification workflows that need to open the scheduling configuration directly.


Category Allocation

This card controls how AI spend is distributed across use-case categories (for example, attack surface management, breach-attack simulation, and similar workloads).

Who can edit: Users with the manage-settings role.

Editing category allocation opens a dedicated modal scoped to allocation percentages only. Changes here do not affect the monthly budget cap, and the modal is accessible independently of super-admin permissions.


AI Spend

This card shows the monthly budget cap and a usage meter. Editing the budget cap is super-admin gated; the edit modal is scoped to the cap only and no longer includes category allocation controls.


Aggressiveness step in the Launch Hannibal wizard

When launching a hunt through the Hannibal wizard, aggressiveness is now a dedicated step rather than an inline selector. Each level is presented as a card with a plain-language description so operators can make an informed choice before proceeding.

Level

Behaviour

Cautious

Lower-impact techniques; reduced risk of service disruption.

Balanced

Standard technique coverage; moderate impact profile.

Aggressive

Broader technique coverage; higher potential for service impact.

Review the descriptions on each card before selecting a level. The selection applies to the hunt being launched and does not change the default for future hunts.


Permissions reference

Capability

Required role

View AI settings page

Any authenticated user with access to the organisation

Edit schedule (toggle and profile)

manage-settings

Edit category allocation

manage-settings

Edit monthly budget cap

Super-admin