Microsoft has published a transparency notice for users of Copilot generative AI products for Microsoft 365, urging companies to ensure they are properly managing AI before deploying it.
The warning comes at a time of growing concerns about data governance, with many companies reluctant to take the plunge until they are confident that AI will protect their interests.
Redmond sums it up this way: “Creating a system that is fit for purpose requires an understanding of how the technology works, its capabilities and limitations, and how to achieve the best performance.”
Microsoft has released Copilot Transparency Note
The accompanimentreleased earlier this week, covers the use of Copilot for Microsoft 365, the enterprise-focused branch of the company’s generative AI efforts. Available to businesses for $30 per user per month, it offers the usual array of GenAI features like summarization, generation, and prediction. However, many businesses are hesitant to give Copilot sensitive corporate information.
In the note, Microsoft describes how Copilot uses large language models (LLMs), including GPT-4, to summarize, predict, and generate content. A user must enter a prompt, and “Copilot for Microsoft 365 sends that prompt to the LLM for processing,” Microsoft explains.
The company also claimed, “The data is encrypted while stored and is not used to train Azure OpenAI Service Foundation LLMs.”
Regarding Copilot’s limitations for Microsoft 365, the tech giant noted that organizations should consider legal and compliance obligations when using AI, especially for organizations in regulated industries.
Microsoft also noted that bias, stereotyping, and unsubstantiated content are potential limitations of the technology: “AI services are fallible and probabilistic.”
While the company notes efforts to improve its services — “Microsoft is investigating the regulatory requirements that apply to Microsoft as a provider of the technology” — the company’s updated Transparency Notice makes clear that companies must put in just as much effort to make AI work for them.