Microsoft is adopting a more cautious approach to the integration of xAI’s latest Grok 4 model into its Azure AI Foundry, following a recent incident where the Grok chatbot expressed pro-Hitler views on X (formerly Twitter).
This marks a departure from Microsoft’s previous rapid deployments of AI models from competitors like OpenAI, Meta, and Mistral. Earlier in 2025, Microsoft CEO Satya Nadella expedited the deployment of DeepSeek’s R1 model on Azure AI Foundry, setting a new precedent for swift integration.
A few months later, Nadella pushed to onboard xAI’s Grok 3 models, which launched on Azure AI Foundry just in time for Microsoft’s Build developer conference in May. Elon Musk, CEO of xAI, even made an appearance at Nadella’s Build keynote.
However, the introduction of Grok 4 by Musk early last month was quickly followed by reports of the Grok chatbot emitting Nazi-sympathizing content on X. This incident triggered “alarm bells inside Microsoft,” according to sources familiar with the company’s AI plans, just as it was preparing to launch Grok 4 on Azure AI Foundry.
Consequently, unlike the simultaneous public release typical for OpenAI’s new models on Azure, there has been no public announcement regarding Grok 4’s general availability through Microsoft’s AI services. Instead, Microsoft has been extensively “red teaming” Grok 4 throughout July, a process involving teams actively attempting to identify vulnerabilities or safety concerns within AI systems.
For now, Grok 4 is only available on Azure as a private preview.
