Microsoft on Monday announced it has added Elon Musk’s controversial Grok AI models to its Azure AI Foundry, becoming the first major cloud provider to offer managed access to the xAI-developed technology.
The decision marks a significant expansion of Microsoft’s AI strategy and opens the door for enterprise customers to use Grok 3 and Grok 3 Mini models with the same service-level agreements applied to other Microsoft-hosted AI products.
“Grok 3 and Grok 3 Mini will have all the service-level agreements Azure customers expect from any Microsoft product,” the company said in a statement, confirming that clients will be billed directly through Microsoft.
Register for Tekedia Mini-MBA edition 17 (June 9 – Sept 6, 2025) today for early bird discounts. Do annual for access to Blucera.com.
Tekedia AI in Business Masterclass opens registrations.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register to become a better CEO or Director with Tekedia CEO & Director Program.
The move gives developers direct, enterprise-grade access to Grok through Azure’s expanding AI Foundry platform, which now offers over 1,900 models, including from Meta, Hugging Face, and other independent developers. It signals Microsoft’s intent to provide customers with a multi-model environment and deeper flexibility, moving beyond its high-profile partnership with OpenAI.
First launched by Musk in late 2023, Grok was marketed as an “edgy,” uncensored alternative to conventional AI assistants. Its behavior often reflected that branding. Grok was found to give answers others wouldn’t, including responding to prompts with vulgar language and engaging in contentious political discourse. Musk, who has accused other AI models of being “woke” and politically biased, positioned Grok as a counterweight.
However, Grok’s has come with its own controversy. A benchmarking study from SpeechMach noted that Grok 3 was among the more permissive models in terms of handling sensitive or potentially offensive content. Recent incidents only intensified scrutiny.
Earlier this year, Grok was found to generate sexualized image manipulations when prompted to undress photos of women. In February, it temporarily suppressed unflattering mentions of Donald Trump and Elon Musk. More recently, the system was reported to repeatedly reference “white genocide in South Africa” following an unauthorized modification.
To address these issues, xAI has taken steps to increase oversight. The company now publishes system prompts for Grok publicly on GitHub and says it has implemented internal changes to prevent unauthorized alterations. The models released through Azure are significantly more locked down than the versions running on X, Musk’s social media platform. According to Microsoft, the Azure-hosted versions include enhanced safety features, enterprise-grade governance, and data integration capabilities that xAI’s own API offering does not currently provide.
The integration also ties into Microsoft’s broader vision for a collaborative AI future. The company has committed to supporting the Model Context Protocol (MCP), an open-source standard introduced by Anthropic that allows different AI agents to work together across platforms. The aim, according to Microsoft’s Chief Technology Officer Kevin Scott, is to build what he calls an “agentic web”—a decentralized network of AI systems that can interoperate much like websites on the internet.
“It means that your imagination gets to drive what the agentic web becomes,” Scott said, “not just a handful of companies that happen to see some of these problems first.”
Microsoft is also experimenting with new approaches to memory retention for AI agents. Rather than relying on computationally expensive large-context models, the company is testing structured retrieval augmentation, a method where agents save short snippets from conversations to create a memory roadmap.
Scott likened the method to how the human brain works: “You don’t brute force everything in your head every time you need to solve a particular problem,” he explained.
The rollout of Grok in Azure positions Microsoft at the center of a rapidly evolving industry where AI tools are becoming more diverse, customizable, and potentially more controversial. Microsoft is betting it can meet enterprise demand without compromising on safety, while also giving a platform to one of the most provocative models in the market, by offering both regulatory guardrails and flexibility.