Microsoft Prioritizes Safety After Rapidly Adopting Generative AI

Since 2017, Microsoft has prioritized AI safety, initiating its efforts with the establishment of the Aether committee to delve into the ethics of artificial intelligence. By 2018, the company had published a comprehensive set of AI principles. However, early 2023 marked a shift as Microsoft rapidly integrated generative AI into its suite of services, including Bing, Edge, Word, and Outlook, leveraging its partnership with OpenAI to accelerate developments.

Now, Microsoft appears to be recalibrating its approach, placing a greater emphasis on safety amid growing concerns from businesses. Recently, the company announced it would cover the legal fees for businesses facing intellectual property lawsuits related to the use of its Copilot products—demonstrating its commitment to responsible AI deployment.

Adelina Balasa, a senior AI cloud solution architect at Microsoft U.K., highlighted the urgency of responsible AI at the Big Data LDN event. She stressed that as companies explore an expanding array of AI products, AI safety must remain a priority. “Generative AI is accelerating tasks and generating content that we can edit, so we are still accountable for the output,” she explained. Balasa emphasized the need for "shared responsibility," indicating that both technology providers and users play crucial roles in ensuring the safety of AI applications.

To support this responsible approach, Balasa advocated for due diligence and adherence to established principles, including those developed by Microsoft. “If you demonstrate your commitment to responsible use, we’re committed to supporting you,” she stated.

### Key Use Cases for Generative AI

One of the most compelling applications of generative AI is the creation of semantic question-and-answer tools, which can significantly enhance data insight exploration. Balasa described these tools as "the next level of enterprise search," addressing a common concern among organizations about the inefficiency of retrieving answers from their own data.

Another major application highlighted by Balasa involves virtual agents. She noted that combining generative AI with traditional technologies, such as speech-to-text, is revolutionizing user experiences. “We’ve evolved from mouse and keyboard to touch screens, and now to voice interaction—all allowing us to engage naturally with these systems,” she said.

### Challenges and Solutions

Alex Karim, Azure Innovation GTM Manager at Microsoft U.K., identified data sovereignty and governance as key challenges in AI implementation. In response, Microsoft has launched a private enterprise version of ChatGPT for Azure, ensuring that user data remains secure. “Your data stays within your Microsoft environment—it doesn’t leave,” Karim assured.

Furthermore, Balasa confirmed that any customer data used within services like ChatGPT for Azure will not be leveraged by Microsoft to refine its models, thus addressing privacy concerns.

In summary, as organizations navigate the complexities of AI deployment, Microsoft is directly addressing safety and governance challenges while promoting responsible practices. The evolution of AI tools and their applications continues to shape the landscape of technology in the enterprise sector.