Microsoft is allowing companies to personalize the AI language model in Azure. This can only be done by sharing company data with the model.
Microsoft is releasing a public preview of ‘on your data’. This is an Azure OpenAI Service in which companies can attach OpenAI language models directly to their data. The data’s location does not matter for the service. It does not have to be placed in the cloud.
According to Andy Beatman, senior product marketing manager at Azure AI, the service gives companies a language model they no longer have to fine-tune themselves. “These models not only leverage their pre-trained knowledge but also access specific data sources, ensuring that responses are based on the latest available information,” he continues.
To get started, you must own an Azure OpenAI Service application and an Azure OpenAI Service resource with the gpt-35 turbo model or the gpt-4 model.
Tip: Microsoft and OpenAI frustrated over each other’s chatbots
Chat Playground
Enterprise data can be linked through the Chat Playground. Here you can choose where the AI language model may browse, looking for an answer. It is possible to link the following locations: Azure Cognitive Search, Azure Blob storage and local files.
Employees can use the personalized AI language model through Azure AI Studio. Otherwise, you can integrate the AI language model into your own applications with Python, cURL, API and SDK.
Secrets public
ChatGPT does however store all information from answers and queries in the Azure OpenAI service. Linking corporate data to the service may therefore pose security risks. There is no guarantee that your corporate data will remain accessible only to your employees and protected from the general public.
Microsoft outlines in its announcement how it is trying to keep the AI language model from harmful reactions and monitors for employee misuse. There is no additional explanation about preventing inadvertent leaks of corporate information.