Target Configuration

Configuring the Target

The final step before enabling the probes is providing the details and capabilities of your target (Figure 1). To do this, fill in all the input fields with your target’s information.

Figure 1: Target Configuration Tab

• Target Name

  • The name of your target that will be displayed within the Platform.

• Target Environment

  • Field for tracking the various stages of your target (Development, Production, and Staging).

• Target Description

  • A brief description of your application’s purpose and use case.

• Target Type

  • You can choose from four target types, depending on whether your application is publicly available or internal, and whether it uses Retrieval-Augmented Generation (RAG) or not.

    • Public With RAG

    • Public Without RAG

    • Private With RAG

    • Private Without RAG

  • Target type determines the default risk priorities for each probe, which are used to calculate your overall risk surface.

• Language

  • Red teaming attacks are generated in English by default, but for better coverage, test variations are also run and the attacks are translated into the other languages selected here.

  • Use this to test multilingual attack inputs against your target.

• Rate Limit

  • The maximum number of requests your application can process per minute.

• Parallel Requests

  • Turned on by default.

  • Toggle off to have the Platform send requests to your target one at a time.

• Multi-Step Attacks

  • Turned on by default.

  • Probe can effectively simulate and test multi-step (multi-message) conversations to evaluate an AI system’s ability to handle context retention, nuanced understanding, and adaptive decision-making across prolonged exchanges.

• Modes Supported

  • This specifies the types of input your application can process, with text set as the default.

  • The Platform also supports attacks through uploaded images, voice, and documents.

  • If your application can handle any of these inputs and you want to test it against multimodal attacks, select the relevant modes.

• RAG File Upload

  • Upload files that are used in your RAG application.

  • Supported formats are: .pdf, .zip, .csv, .txt, .md, .gzip, .xz, .bz2, .docx, .doc, .pptx., .ppt, .xls, .xlsx.

  • Up to 5 MB.

  • Serving as a source for retrieval to provide the model with relevant data before generating responses.

• RAG File Facts Limit

  • Maximum number of facts that can be extracted from the RAG file.

  • Number should be greater than 3 per uploaded RAG file (e.g., If a .zip is uploaded with 3 RAG files inside, number of facts limit should be at least 9).

• Predefined Responses

  • Predefined responses are the expected messages returned by the target in specific situations, for example, when an adversarial input is sent and the system activates a guardrail.

  • If the system’s actual response matches a predefined response during an adversarial testing scenario (e.g., the system replies with “message is blocked”), the probe test case is marked as passed.

  • Predefined responses can be defined in two ways:

    • Text: A direct, 1:1 match. The system response must be exactly identical to the predefined message.

    • Regex: A pattern-based match, defined using standard regular expression rules, allowing for flexible response validation.

Supported attack translation Languages (Default Variation Attacks)

• Albanian

• Arabic

• Armenian

• Azerbaijani (Latin)

• Bosnian (Latin)

• Bulgarian

• Catalan

• Chinese

• Chinese Simplified

• Croatian

• Czech

• Danish

• Dutch

• English

• Estonian

• Faroese

• Fijian

• Filipino

• Finnish

• French

• French (Canada)

• Georgian

• German

• Greek

• Haitian Creole

• Hebrew

• Hindi

• Hungarian

• Icelandic

• Igbo

• Indonesian

• Irish

• Italian

• Japanese

• Kazakh

• Korean

• Latvian

• Lithuanian

• Macedonian

• Maltese

• Nepali

• Norwegian Bokm

• Persian

• Polish

• Portuguese

• Romanian

• Russian

• Serbian (Cyrillic)

• Serbian (Latin)

• Slovak

• Slovenian

• Spanish

• Swedish

• Thai

• Tibetan

• Turkish

• Turkmen (Latin)

• Ukrainian

• Uzbek (Latin)

• Vietnamese

• Welsh

Saving the Target

After entering all the required inputs, your target can be saved, and you can proceed to configure the probes. Target is saved by clicking the "Save" button.

A success notification will indicate that your new target has been saved successfully. Your target will be automatically selected, and its name will appear in the targets list within the drop-down menu. The Probe Settings page will be displayed, allowing you to start configuring the probes.

If you need to make any changes later, you can edit the target at any time on the Target Settings page.