As AI advances, we all have a role to play in unlocking the positive impact of AI for organizations and communities around the world. That is why we are focused on helping customers use and create AIreliable, that is, an AIsafe, protectedandprivate.
At Microsoft, we are committed to ensuring trustworthy AI and are developing cutting-edge support technology. Our commitments and resources go hand in hand to ensure that our clients and developers are protected at all layers.
Building on our commitments, we are today announcing newresourcesof products to strengthen the safety, security and privacy of AI systems.
Security.Security is our top priority at Microsoft, and ourSecure Future Initiative(SFI) underscores our company-wide commitments and the responsibility we feel to make our customers moreinsurance. This week, we announced our firstSFI Progress Reporthighlighting updates that cover culture, governance, technology, and operations. This fulfills our promise to prioritize safety above all and is guided by three principles: secure by design, secure by default, and safe operations.In addition to our primary offers,Microsoft Defender e PurviewOur AI services have basic security controls, such as internal functions to help prevent immediate injections and copyright violations. Based on them, today we are announcing two new features: In addition to our primary offerings,Microsoft Defender e PurviewOur AI services have basic security controls, such as internal functions to help prevent immediate injections and copyright violations. Based on them, today we are announcing two new features:
- Reviews no Azure AI Studio to support proactive risk analysis.
- Microsoft 365 Copilot will providetransparency in web queriesto help administrators and users better understand how web search enhances Copilot's response.Available soon.
Our security features are already being used by customers.A Cummins, a 105-year-old company known for its engine manufacturing and development of clean energy technologies,turned to Microsoft Purviewto strengthen your data security and governance by automating data classification, tagging, and labeling.A EPAM Systems, a software engineering and business consulting company, implemented o Microsoft 365Copilot for 300 users due to the data protection they receive from Microsoft. J.T. Sodano, Senior IT Director, shared that "we were much more confident with Copilot for Microsoft 365 compared to other LLMs (large language models) because we know that the same data and information protection policies we set up inMicrosoft Purviewthey apply to Copilot.
Security.Including security and privacy, the broader principles ofMicrosoft Responsible AI, established in 2018, continue to guide how we create and implement AI safely across the company. In practice, this means building, testing, and properly monitoring systems to prevent undesirable behaviors such as harmful content, bias, misuse, and other unintended risks. Over the years, we have made significant investments in building the governance framework, policies, tools, and processes necessary to uphold these principles and to build and deploy AI safely. At Microsoft, we are committed to sharing our learnings on this journey of advocating for our Responsible AI principles with our customers. We use our own best practices and learnings to provide people and organizations with resources and tools to create AI applications that share the same high standards we strive for.
Today, we’re sharing new capabilities to help customers realize the benefits of AI while mitigating the risks:
- A resourceof correctionin the grounding detection feature of Microsoft Azure AI Content Safety that helps correct hallucination issues in real-time before users see them.
- Embedded Content Security, allowing customers to embed Azure AI Content Safety into devices. This is important for device scenarios where cloud connectivity may be intermittent or unavailable.
- New reviewsin Azure AI Studio to help clients assess the quality and relevance of outputs and how often their AI application generates protected material.
- Detection of protected material for programmingIt is now in preview version on Azure AI Content Safety to help detect existing content and code. This feature helps developers explore public source code on GitHub repositories, promoting collaboration and transparency, while also enabling more informed coding decisions.
It's amazing to see how clients from all sectors are already using Microsoft solutions to create safer and more reliable AI applications. For example, a Unity,a platform for 3D games, used Microsoft Azure OpenAI Service to create Muse Chat, an AI assistant that facilitates game development. Muse Chat uses content filtering models inAzure AI Content Safetyto ensure the responsible use of the software. Furthermore, theASOS,a UK-based fashion retailer with nearly 900 brand partners, used the same built-in content filters in Azure AI Content Safety to support high-quality interactions through an AI application that helps customers find new looks.
We are also seeing the impact in the educational space.New York City Public SchoolsThey partnered with Microsoft to develop a secure and appropriate chat system for the educational context, which they are now testing in schools.The South Australian Department of Educationalso brought generative AI to the classroom with EdChat, relying on the same infrastructure to ensure safe use for students and teachers.
Privacy.Data is at the heart of AI, and Microsoft's priority is to help ensure that customer data is protected and compliant through ourprivacy principleslong-standing data, including user control, transparency, and legal and regulatory protections.To develop this, today we are announcing:
- Confidential inferencein preview versionin our Azure OpenAI Service Whisper model, so that clients can develop generative AI applications that support end-to-end verifiable privacy.The confidential inferenceIt guarantees that the client's confidential data remains secure and private during the inference process, which is when a trained AI model makes predictions or decisions based on new data. This is especially important for highly regulated sectors such as healthcare, financial services, retail, manufacturing, and energy.
- General availability of Azure Confidential VMscom GPUs NVIDIA H100 Tensor Core, allowing customers to protect data directly on the GPU. This is based on our confidential computing solutions, which ensure that customer data remains encrypted and protected in a secure environment so that no one has access to the information or system without permission.
- Azure OpenAI Data Zonesfor the European Union and the United States are arriving soon and are based on the existing data residency provided by the Azure OpenAI Service, facilitating the management of processing and storage of generative AI application data. This new feature offers customers the flexibility to scale generative AI applications across all Azure regions within a geography, while giving them control over data processing and storage within the EU or the US.
We have seen growing customer interest in confidential computing and enthusiasm for confidential GPUs, including from application security providerF5, which is using Azure confidential VMs with NVIDIA H100 Tensor Core GPUs to create advanced AI-based security solutions, ensuring the confidentiality of the data that your models are analyzing. And the multinational banking corporationRoyal Bank of Canada (RBCintegrated Azure confidential computing into their own platform to analyze encrypted data, preserving customer privacy. With the general availability of Azure confidential VMs with NVIDIA H100 Tensor Core GPUs, RBC can now leverage these advanced AI tools to work more efficiently and develop more powerful AI models.
Achieve more with trusted AI
We all need and hope for an AI we can trust. We saw what is possible when people are empowered to use AI reliably, fromenrich employee experiencesandremodel business processesuntilreinventing customer engagementandreimagine our liveseveryday life. Withnew featuresthat improve security, protection, and privacy, we continue to enable customers to use and create reliable AI solutions that help all people and organizations on the planet achieve more.Ultimately, trustworthy AI encompasses everything we do at Microsoft and is essential to our mission, as we work to expand opportunities, build trust, protect fundamental rights, and promote sustainability in everything we do.
