# AI Attack Simulation for Custom Applications

AI attack simulations, as well as system prompt evaluations, can be run within the HiddenLayer console against a variety of foundation model providers (see  AI Attack Simulation Overview (Console) for more information).

However, a large part of the value of these tools lies in the possibility of configuring attacks against custom applications, not just foundational models. Attack simulations and system prompt evaluations can also be performed via API against any application that is accessible through the internet. This section will show you how to configure those simulations using the HiddenLayer SDK.

Accessing Applications
AI Attack Simulation is only available as a SaaS offering from HiddenLayer. In order to test custom applications, those applications must have a publicly exposed interface (endpoint, widget, etc.) such that they can be reached by our "attacker", and must be able to return a response back to the testing endpoint.

Running simulations against your custom applications programmatically allows you to:

* Establish a security baseline
* Identify vulnerabilities in application stacks, not just in their component parts
* Track and compare quantifiable security metrics
* Integrate evaluations and simulations as part of the development and release pipeline


This approach helps secure AI applications throughout their development lifecycle; allows teams to create a baseline for acceptable security standards; and ensures that application security remains high throughout the application's life.