Microsoft launches tools for more security and discovery in Azure AI – IT Pro – News

Microsoft is launching a set of new tools for AI developers running their software on Azure. This includes a tool to instantly detect injection attacks, monitor risks, and recognize hallucinations in the output.

The tools will be available for Azure AI Studio, Microsoft writes. Developers who run their AI programs in Azure AI can use it. The company says there is a growing demand for tools that help “balance innovation and risk management.” The new tools for Azure AI Studio should help with this.

Microsoft releases five such tools. Demands armor It is a mechanism to detect and stop jailbreaks and other forms of injection attacks. Many users of AI systems try to “hack” them by going beyond the limits of what these tools allow. This can be done, for example, via jailbreaks such as Then for ChatGPT, but also by seeing how generative AI responds to specific code or by uploading documents. Prompt Shield looks at commonly used methods and common phrases by which this is done, or which methods are often used in jailbreaking. Then blocking these methods becomes easier.

Azure AI Studio goes further, too Detection of hallucinations in output Generative artificial intelligence. This is particularly dangerous in areas where data, including outputs, are best kept confidential. When models are hallucinating, such a model can leak data that it is not actually supposed to leak. “Ground detection” prevents this. There will also be one Risk and safety monitoring toola kind of platform where admins can see the impact of content filters at a single glance.

Finally, there will be two tools that can make generative AI outputs more secure. Safety messaging system It can automatically impose safety conditions on the output. Safety assessments This program aims to discover a security vulnerability in the application against jailbreak attacks, as well as assess the risks of the content.

Not all tools are immediately available. Real-time shields, safety assessments, risk monitoring, and safety are in preview in Azure AI and Azure OpenAI, but floor detection systems and safety messaging are coming “in the future.”