Microsoft’s Responsible AI Transparency Report, which details the company’s initiatives to introduce responsible AI platforms, was recently published. The study mostly discusses Microsoft’s efforts to securely implement AI products in 2023.
The corporation claims in the report that it developed thirty responsible artificial intelligence capabilities in the previous year.
The paper also highlights the growth of Microsoft’s responsible AI team and the deployment of strict risk assessment methods for teams developing generative AI products at every development step.
According to Microsoft, it added a watermark to an image identifying it as created by an AI platform by integrating Content Credentials into its image generating systems.
Along with tools for evaluating security threats, Microsoft has also given Azure AI clients improved tools for identifying problematic content, such as hate speech, sexual content, and self-harm.
The business is expanding on its red-teaming efforts, which now include external red-teaming apps that enable independent assessments prior to model releases, as well as internal red teams that are tasked with stress-testing safety features in its AI models.