We are thrilled to announce the public preview of hybrid Azure AI Content Safety (AACS), which will cater to organizations seeking robust, flexible content safety mechanisms both in the cloud and on-device. 

What is Hybrid AACS? 

  
The Hybrid AACS is an innovative solution that includes a connected or disconnected container for on-premises (customer’s data center consumption and an embedded SDK for on-device processing. It empowers organizations to implement state-of-the-art safety measures, ensuring content is managed securely and efficiently, regardless of the environment. 

Connected or disconnected Container: Utilizing containers in your data centers and environments ensures that sensitive data remains under your control, reduces data processing latency, and helps meet specific security and data governance requirements. Furthermore, Disconnected containers enable you to use several APIs disconnected from the internet. 

Embedded SDK: Facilitates real-time content safety checks directly on device, ideal for scenarios where internet and cloud connectivity are intermittent or unavailable. 

Why Hybrid AACS? 

 
Enterprises handle increasingly sensitive information across a variety of platforms. In the era of generative AI (GenAI), where AI is becoming more pervasive there is need for comprehensive AI safety solutions. Our new Azure AI Content Safety solutions – Embedded Content Safety, and Content Safety Container –aim to fill this gap by providing below values: 

Greater Control: With on-device processing and in-house cloud capabilities, customers have control over data flows, providing peace of mind for organizations handling sensitive information. 

Flexibility and Scalability: Organizations can tailor the deployment based on their specific needs, easily scaling up as their requirements grow. 

Reduced Latency: Immediate processing on devices or within a local environment minimizes delays, enhancing operational efficiency. 

Less demand on network connectivity: device scenarios don’t always have good network connectivity, when it becomes unstable, cloud traffic can safely fall back to on-device AI content safety SDK. 

https://youtu.be/YrEeonwWZJU?si=jlwCS5OAiTnOq8_K&t=962  

How does it work? 

  
The Hybrid AACS operates in two parts: 
 

Connected or disconnected Container for on-prem consumption: We will release content safety containers, including connected containers and disconnected containers for text and image models. The containers allow enterprises to deploy AI content safety features within their own infrastructures, ensuring that sensitive data remains within their control and reducing latency in data processing. 
 

Embedded SDK for On-Device AI Safety: We will release a C++ SDK with embedded content safety model. The embedded model is also optimized to run on devices which have less powerful computing resource compared to Azure high end GPUs. In this release, the embedded content safety can run on Windows PC devices. We plan to expand to more platforms and devices in future release.  This functionality empowers real-time content safety checks directly on devices, ensuring all data remains on the device. It also significantly reduces the computation needed for model inferencing, ensuring efficient performance even on smaller devices. 

  
Performance Evaluations 

We’ve conducted performance benchmark tests on various CPUs and GPUs to help you determine if your device is suitable for running embedded content safety. For detailed performance data and SDK parameters that can impact performance, refer to our Performance Benchmark Data. 

Getting Started 

  
To get started with Hybrid AACS, visit the Azure documentation website for detailed information on deployment options, technical requirements, and support resources. Azure also provides customer service and technical support to assist with implementation. 
 
For more detailed information, please refer to below documentation 

Embedded Content Safety Documentation  

Content Safety Containers Overview. 
 

We hope you are as excited as we are about the release of Hybrid Azure AI Content Safety! 

​Microsoft Tech Community – Latest Blogs –Read More