How AI Content Safety Protects AI Systems from Bad Actors

• Azure AI Content Safety provides several capabilities intended to safeguard generative AI applications from malicious use.
• The services are new and developing rapidly; organizations will need to adjust and evolve their defenses as attackers become more sophisticated.
• Organizations will need overlapping approaches including monitoring, internal testing, and the ability to respond rapidly should incidents occur.

Azure AI Content Safety is a set of services that aim to help developers protect their generative AI applications and the underlying large language models (LLMs) from misuse. These services are available as APIs for developers and also from within Azure AI Foundry. When used as an API, they identify potential problems, and it is up to the developer to decide how to proceed. Azure AI Foundry provides developers with a UI and the ability to automatically block content based on the results.

This report, the second in a series on AI Content Safety, explains the capabilities offered by Azure AI Content Safety as well as their limitations. The full series of reports includes the following: