IBM launches DataPower Interact Gateway for governed AI interactions

IBM has announced DataPower Interact Gateway, a runtime infrastructure product for governing AI interactions where enterprise applications connect to models, tools, APIs and business systems.

A gateway for AI traffic

In its announcement, IBM describes DataPower Interact Gateway as an “AI mediation governance gateway” intended to secure, govern and observe interactions among AI agents, large language models, tools, APIs and enterprise systems. The company positions the product as part of its DataPower portfolio, extending familiar gateway controls into AI application environments.

IBM’s product page says DataPower Interact Gateway is designed to mediate traffic for both LLM access and Model Context Protocol interactions. In practical terms, IBM says the gateway can sit between enterprise applications and model providers, tools or APIs, applying policy controls and collecting analytics as those interactions occur.

The product is available now. IBM Support lists DataPower Interact Gateway 12.1 as generally available on June 26, 2026.

LLM and MCP governance

IBM Documentation says DataPower Interact Gateway can function in two primary roles: as an MCP Gateway and as an LLM Gateway.

As an MCP Gateway, IBM says the product can expose APIs as tools, allowing AI applications to use enterprise services through governed interfaces. As an LLM Gateway, IBM says it provides governed access to model-provider APIs, placing enterprise controls between applications and the external or internal models they call.

That distinction is important because enterprise AI systems increasingly involve more than a single chatbot connected to one model. Applications may call multiple models, retrieve business data, invoke APIs, or use tools that perform actions in enterprise systems. IBM’s framing suggests DataPower Interact Gateway is meant to provide a control point across those interactions rather than leaving each application team to implement governance separately.

Observability, usage and cost controls

IBM’s product page also highlights analytics for usage, performance and cost control. According to IBM, the gateway can provide visibility into AI interaction patterns, including how services are being used and how they are performing.

Those capabilities address a growing operational challenge for companies adopting generative AI: model usage can be distributed across many applications, providers and teams. A gateway approach gives enterprises a central place to observe traffic, enforce access policies and manage consumption.

IBM’s announcement emphasizes security and governance rather than model development. The product is presented as runtime infrastructure for enterprises that are already connecting AI systems to business applications, data sources and APIs.

Part of IBM’s enterprise AI infrastructure push

DataPower has long been associated with enterprise API and integration security. With DataPower Interact Gateway, IBM is extending that gateway model to AI interactions, including LLM calls and MCP-based tool access.

IBM’s documentation and product materials do not present the gateway as a model provider or an AI application builder. Instead, they describe it as infrastructure for controlling the interactions that happen when AI applications communicate with models, tools and enterprise systems.

For organizations experimenting with AI applications, that may be a secondary concern. For companies moving AI workloads toward production, IBM is arguing that the interaction layer itself needs governance: who can call which models, which APIs can be exposed as tools, how usage is monitored, and how performance and cost are managed.

The release gives IBM a dedicated product for that layer. DataPower Interact Gateway 12.1 is now listed by IBM as generally available, with official documentation covering management of AI services through both MCP Gateway and LLM Gateway functions.