AI Usage and Limits

UTMStack v12 can limit how many AI requests each tenant makes per day. This page explains what counts as a request, where the limit comes from, and where you see the day's usage.

What counts as one request

Each of these counts as one AI request for the tenant:

  • A message you send in the SOC-AI assistant.

  • A click on Generate AI summary in an alert.

  • One alert analysed by automatic triage (Auto-analyze alerts).

Requests are counted per tenant and per day. The day resets at midnight UTC.

When there is no limit, requests are not counted.

Where the limit comes from

Two values decide a tenant's daily allowance:

ValueWhere it is setDefault
Instance limitA setting of the UTMStack server.No limit.
AI request cap of the tenantTenants page, edit the tenant, switch on Cap this tenant ("An operational cap, per day. Uncapped, they spend against the instance licence."). Only on a multi-tenant installation, by the instance operator.No cap.

A tenant never gets more than the instance limit. If the tenant has no cap of its own, the instance limit applies.

What the number means:

ValueMeaning
No valueNo limit of its own; the instance limit applies.
0The AI is not available to this tenant.
A number, for example 200That many requests per day.

On a single-tenant installation there is no Tenants page, so in practice there is no daily limit.

When the allowance is used up

Further requests that day are refused with the message "this tenant has used the AI requests its plan allows for today". The assistant cannot answer and Generate AI summary fails until midnight UTC. Automatic triage skips new alerts for the rest of the day: they arrive without a SOC AI assessment, and they are not analysed later.

Over the API, the same refusal is HTTP 429 with the fields limit and used (see SOC-AI API).

Where you see usage

The SOC-AI Configuration page showing AI allowance today with the requests used out of the daily allowance, and the Usage block with Chat completions (today) and when the count resets
PlaceWhat it shows
Configure > Settings > SOC-AI, AI allowance todayRequests used today "of N requests", or "requests — the AI is not available to this tenant". Hidden when there is no limit. The hint reads "Set by whoever runs the instance. It resets at midnight UTC."
Tenants page, AI todayEach tenant's usage today, or "no cap".
Configure > Settings > SOC-AI, Usage (instance operator using ThreatWinds)Chat completions (today) at ThreatWinds and "Resets in …".

Programs can read the same numbers with GET /api/v1/soc-ai/usage, which answers with limit, used, remaining (only when a limit is set) and resetsAt (UTC time). See SOC-AI API.

ThreatWinds billing

With the ThreatWinds provider, AI use is "Billed through your ThreatWinds subscription." With any other provider, your provider bills you directly under its own terms.

  • AI Providers

  • SOC-AI Assistant