Run GitLab's AI Coding Assistant in Your Cloud

GitLab's AI assistant, Duo, now supports self-hosting on Microsoft Azure, giving companies total control over their code and AI data.

GitLab is expanding its AI capabilities to give enterprises more control over their data, according to a report from InfoQ. The company announced that GitLab Duo Self-Hosted, its suite of AI-powered development tools, now supports models deployed through Microsoft's Azure AI Model Foundry. This integration allows organizations to run GitLab's AI features, such as code suggestions and vulnerability summaries, using large language models that are hosted entirely within their own private Azure environment. For companies concerned about sending sensitive source code to external services, this move directly addresses a major barrier to adopting AI in the software development lifecycle, placing control over data privacy and security squarely in the hands of the customer.

The new capability works by rerouting AI requests from a self-managed GitLab instance to a model deployed within the customer's Azure subscription. Instead of GitLab Duo communicating with a third-party API, it connects to an AI model that the organization has selected and deployed via Microsoft's Foundry service. Microsoft Foundry acts as a curated catalog, allowing businesses to access and run powerful models like Meta's Llama family or Mistral's offerings on their own dedicated Azure infrastructure. This ensures that code snippets, prompts, and other proprietary data used to generate AI suggestions never leave the company's secure network perimeter. The connection is managed through GitLab, but the data processing and model inference happen in the customer's controlled cloud environment, providing a transparent and auditable alternative to typical SaaS-based AI tools.

This partnership reflects a crucial trend in the enterprise AI market: the shift from public, multi-tenant AI services to private, single-tenant deployments. While AI coding assistants offer immense productivity gains, many large enterprises, especially in regulated industries like finance, healthcare, and defense, have been hesitant to adopt them due to data privacy, intellectual property, and compliance risks. The fear of proprietary code being used to train a public model or being exposed through a breach is a significant concern for CTOs and Chief Information Security Officers. This "bring your own AI" model, where companies can use their preferred infrastructure to run vetted models, is emerging as the dominant solution. It allows businesses to balance the benefits of generative AI with their stringent security postures, a model also being pursued by competitors like GitHub with its Copilot Enterprise offerings.

For technology leaders, this development makes GitLab Duo a much more viable option for internal development teams. The ability to self-host the AI backend on a trusted cloud platform like Azure removes a major adoption blocker and puts GitLab on a stronger competitive footing. Security and IT teams can now apply their existing Azure security policies, monitoring, and access controls to the AI workload, treating it like any other sensitive internal application. This simplifies compliance with regulations such as GDPR, HIPAA, and CCPA. Looking ahead, the key question is how GitLab will expand this model. Observers should watch for similar integrations with other major cloud AI platforms, such as Amazon Web Services' Bedrock and Google Cloud's Vertex AI. A multi-cloud self-hosting strategy would further cement GitLab's position as a flexible, platform-agnostic tool for enterprise DevOps and DevSecOps.

For development and security teams, this integration is a critical enabler for adopting AI tools in regulated industries. It means they can use GitLab Duo's features without sending proprietary code to third-party services, satisfying strict data residency and compliance requirements that previously blocked AI adoption.

This move allows businesses to leverage AI-powered development without compromising on data security or regulatory compliance. It reduces the risk of intellectual property leakage and can accelerate software delivery by unblocking teams that were previously prohibited from using cloud-based AI coding assistants.