Microsoft has recently announced the launch of Azure AI Content Safety, a sophisticated service designed to empower businesses with the capability to identify and filter out potentially harmful content across a spectrum of applications and services.
Good candidates for this are:
Businesses wishing to ensure compliance with regulations or maintain the intended environment for their users in M365
Online marketplaces that moderate product catalogues and other user-generated content.
Gaming companies that moderate user-generated game artefacts and chat rooms.
Social messaging platforms that moderate images and text added by their users.
Enterprise media companies that implement centralised moderation for their content.
Education solution providers filtering out content that is inappropriate for students and educators.
The service encompasses both text and image recognition, effectively flagging content that Microsoft classifies as "offensive, risky, or undesirable." It can detect profanities, adult content, violence, gore, and specific forms of problematic speech.
Louise Han, the product manager for Azure Anomaly Detector, conveyed in a blog post that Azure AI Content Safety is a step towards creating a safer digital environment that encourages the responsible use of AI while safeguarding the well-being of individuals and society.
Azure AI Content Safety is versatile, proficient in handling diverse content categories, languages, and potential threats in moderating textual and visual content. It features image recognition capabilities that utilize AI algorithms to scan, analyse, and moderate visual content, thereby ensuring what Microsoft terms "360-degree comprehensive safety measures."
Furthermore, this service is equipped to moderate content in multiple languages and employs a severity metric that quantifies the degree of concern for specific content on a scale from 0 to 7.
Content graded 0-1 is considered safe for all audiences, while content rated 2-3 is characterized as low-level, expressing prejudiced, judgmental, or opinionated views.
Content with medium severity falls within the 4-5 range, denoting offensive, insulting, mocking, or intimidating language, or explicit attacks against identity groups. Content rated 6-7 is deemed high severity, indicating content that promotes harmful acts or endorses extreme forms of harmful activity towards identity groups.
Azure AI Content Safety also implements multicategory filtering to identify and categorize harmful content across various critical domains, encompassing hate speech, violence, self-harm, and sexually explicit material.
In an era where AI-generated content is becoming increasingly prevalent, Azure AI Content Safety aims to extend its protective reach beyond just human-generated content. By ensuring the accuracy, reliability, and absence of harmful or inappropriate materials in AI-generated outputs, this service not only safeguards users from misinformation and potential harm but also upholds ethical standards and fosters trust in AI technologies.
Azure AI Content Safety is now generally available (GA). If your interested in implementing this, please give us a call.
Comments