Microsoft Introduces AI Safety Ranking on Azure

Microsoft Corporation has introduced a new safety ranking feature for AI models on its Azure Foundry platform, according to the Financial Times. This initiative is designed to address increasing concerns about data protection within AI products.

The safety ranking will be added to Microsoft's existing "model leaderboard," which evaluates AI models based on quality, cost, and throughput. This new category will allow cloud customers to assess AI products from various providers, including OpenAI and China's DeepSeek, based on their safety performance.

Sarah Bird, Microsoft's head of Responsible AI, stated that the safety ratings will be based on benchmarks like Microsoft's ToxiGen, which detects hate speech, and the Center for AI Safety's Weapons of Mass Destruction Proxy benchmark. These metrics aim to help customers make informed decisions about which AI models best meet their trust and risk requirements.

The leaderboard is accessible to tens of thousands of clients using the Azure Foundry developer platform and is expected to influence the demand for AI models and applications. This move underscores Microsoft's commitment to enhancing data security and privacy for its cloud customers.