Microsoft has released a Transparency Note for those using its Copilot for Microsoft 365 generative AI products, urging businesses to ensure they’re managing AI correctly before deploying it.
The warning comes amid growing concerns over data governance, leaving many companies reluctant to take the plunge until they can be sure that AI is protecting their interests.
Redmond summarized: “Creating a system that is fit for its intended purpose requires an understanding of how the technology works, what its capabilities and limitations are, and how to achieve the best performance.”
Microsoft issued Copilot Transparency Note
The guidance, issued earlier this week, covers the use of Copilot for Microsoft 365, the business-focused arm of the company’s generative AI efforts. It’s available to businesses paying $30 per user per month, and offers the usual array of GenAI functionalities like summarization, generation and prediction. However, many companies have been hesitant to feed Copilot sensitive company information.
In the Note, Microsoft details Copilot’s use of large language models (LLMs), including GPT-4, to summarize, predict and generate content. A user must enter a prompt, then “Copilot for Microsoft 365 sends this prompt to the LLM for processing,” explains Microsoft.
The company also asserted: “The data is encrypted while it’s stored and isn’t used to train Azure OpenAI Service foundation LLMs.”
Speaking about the limitations of Copilot for Microsoft 365, the tech giant stated that organizations should consider legal and compliance obligations when using its AI, particularly for those in regulated industries.
Microsoft also noted that bias, stereotyping and ungrounded content are potential limitations of the technology: “AI services are fallible and probabilistic.”
While the company makes comments about efforts to improve its services – “Microsoft is examining regulatory requirements that apply to Microsoft as a provider of the technology” – the company’s updated Transparency Note makes it clear that businesses much put in the same amount of effort to ensure that AI works for them.
More from TechRadar Pro
