Ignition and AI
Ignition is a commercial off the shelf (COTS) software platform delivered under an on-premises shared responsibility model: Inductive Automation produces and supports the software; you install it, run it on your own infrastructure, and control connections. The same applies with AI.
Inductive Automation does not develop or host AI models, does not operate any hosted service that processes your data, and never sees what your Ignition system sends to your model providers.
Current releases (Ignition 8.1 and 8.3)
Ignition 8.1 and 8.3 contain no native AI or machine learning features. Customers who want AI or ML in their systems connect to external tools through Ignition's standard integration capabilities (scripting, APIs, databases, and modules). Those connections are built, configured, and governed entirely by the customer. This often looks like Ignition supplying process data to external data repositories, where 3rd party AI is separately used.
Ignition 2027
Beginning with Ignition 2027, Ignition will offer optional AI capabilities. They are delivered as modules, are off until you turn them on, and can be disabled or removed entirely.
Catalyst Core (included with platform) helps engineers design and operate Ignition itself. It provides contextual assistance inside the Designer (scripts, expressions, queries, bindings) and AI generated status summaries in the Gateway web interface. Designer assistance only proposes changes; nothing is written to a project until a user reviews and accepts it.
Catalyst Agent (sold separately) lets customers build AI powered functionality into their own Ignition applications. Customers define agents (model, instructions, permissions), author the tools those agents can call using ordinary Ignition scripting, and expose them to users through scripting or a Perspective chat component. Agents can only act through tools the customer has written and are bound by Ignition's existing role and security zone model.
What this means for your security / risk posture
You choose the model, or none at all. Ignition does not ship with a model. AI features work only after an administrator configures a model provider. Supported options include commercial providers (such as Anthropic and OpenAI) and self hosted models (such as Ollama) for sites where data cannot leave the network. Model choice, data handling, and the terms of any model provider relationship are entirely yours.
Opt in at every layer. The AI modules are separately installed and licensed, present an acceptance gate before first use, and are governed by Gateway settings that an administrator controls. No AI feature activates on its own.
Your data stays under your control. Any data sent to a model goes directly from your Gateway or Designer to the provider you configured. Inductive Automation is not in the path and receives nothing. Provider API keys are stored using Ignition's existing secrets management.
Existing security controls apply. Access to agents and Model Context Protocol servers is governed by the same role based and security zone permissions used throughout Ignition. Session limits, timeouts, and enable/disable controls are available per agent and per server.
Development practices
The AI modules are developed under the same secure development lifecycle as the rest of Ignition, certified to IEC 62443-4-1 (ISASecure SDLA) and aligned with NIST SSDF, and are in scope for product penetration testing.
This page will be updated as the Ignition 2027 release approaches with final module details, scope, and configuration details.