Incident AI that never leaves your network

Production logs are some of the most sensitive data an organisation holds. They carry account numbers, internal hostnames, customer identifiers and the exact shape of the systems an attacker would most like to map. For many teams, sending them to an outside AI service is not a risk to be weighed. It is simply not allowed.

That rules out most AI tooling for incident work, and it is why Pindot was built to run where the data already lives. This post explains what that means in practice.

The question every security team asks first

When we talk to security teams, the first question is almost always the same: does any of our data leave our network? For an on-premise Pindot, the answer is no. Everything runs inside your network, the AI included. The only way anything goes out is through a destination you switch on yourself, such as Jira or Slack, and an air-gapped installation can keep even those inside your own walls.

What runs where

• The product runs on servers you own and operate, alongside the systems it reads.
• The AI runs on your hardware too. Questions and answers are never sent to an outside model.
• Sources are read with read-only access through accounts you create and control.
• Findings leave only through destinations an administrator has switched on.

Fully air-gapped, when you need it

Some environments are sealed off from the internet entirely. Pindot needs no outside connection to install, run or answer, so it works the same way there. Updates arrive as offline packages that your team applies on its own schedule, and nothing phones home. Findings can still reach the team, through tools that live inside the same network.

What stays the same wherever it runs

Some organisations would rather not operate the platform themselves, and for them Pindot can be hosted in a private environment run by us, reaching your systems over an encrypted connection. Where it runs changes who operates it. It does not change what it does, or how it protects your data.

• Read-only, always: it reads your systems and never writes to them.
• Evidence with every answer: each answer cites the exact records behind it.
• A full audit trail: every question is recorded, and the record is never edited.
• Access you control: sources are granted per team and per person, and the AI is only shown what the person asking is allowed to see.

Choosing between the options

The choice usually follows from two questions. Must the data stay inside your own network, by policy or by regulation? Then on-premise is the answer, air-gapped if the network is sealed. Would you rather not run the platform yourselves, and is an encrypted connection to a private environment acceptable? Then hosted is the faster start. Many teams begin with one and revisit the decision later, and our deployment page compares the two side by side.

Bring your security questionnaire

Every environment has its own rules, and the details matter more than any summary. Bring your security questionnaire to a demo and we will answer it with you, line by line, including exactly what Pindot needs from each system it reads.