Microsoft Azure OpenAI Service now offers real time detection of abusive requests

In recent months, there have been reports of users successfully using generative AI services to bypass content restrictions and create content. One notable instance occurred earlier this year when explicit deepfake images of pop artist Taylor Swift circulated on social media. These images were allegedly produced by a group utilizing specific text prompts in the AI image generator Microsoft Designer.

Today, Microsoft unveiled a public preview of a new feature available on its Azure OpenAI Service. This service enables businesses and organizations to utilize OpenAI’s generative AI models through Microsoft’s Azure cloud servers. The recently introduced “Risks & safety monitoring”feature aims to provide businesses with a live overview of their AI systems’ usage and identify any attempts of misuse by individuals.

According to a recent blog post, Microsoft has introduced a feature that can display data about various end-user requests on their AI services which are deemed to be unsafe. These requests encompass:

• Total blocked request count and block rate
• Blocked requests distribution by category
• Block rate over time by category
• Severity distribution by category
• Blocklist blocked request count and rate

Using this information, businesses can modify their AI content guardrails to ensure that their services are not utilized in a manner that goes against their guidelines.

Additionally, there is a function that assists businesses in identifying users who repeatedly submit abusive or detrimental requests to their AI services. Microsoft states:

If any content from a user is flagged as harmful and combining the user request behavior, the system will make a judgement on whether the user is potentially abusive or not. Then a summarized report will be available in the Azure OpenAI Studio for further action taking.

The company has the ability to modify its system in accordance with its own set of rules and regulations in order to safeguard against potential breaches caused by these troublesome end users.

The recently added monitoring capabilities for Azure OpenAI Service can be accessed by customers in the East US, Switzerland North, France Central, Sweden Central, and Canada East regions. It is currently unclear when this public preview will be extended to additional countries or become generally available.