The test details report provides analysis and results from each test so you can leverage the report to see the overall result of the test round and make adjustments to your organization’s AI usage policies.
To view the test details:
-
Go to AI Red Teaming.
-
Click the Test Rounds tab.
-
Click the Name of the test you want to view the report for.
Test Details Report
In the Test Details report, you can see the following information:
General
-
Test Name: The name of the test.
-
Target: The registered AI target that the test was performed on.
-
Language: The language used for the test round evaluation.
-
Executed at: The date when Netskope performed the test.
-
Test ID: The ID of the test.
-
Status: The status of the test.
-
Completed: The test is complete.
-
Ongoing: The test is still in progress.
-
Errored: The test was automatically cancelled due to a large number of errors.
-
-
Note: Optionally, enter any additional comments or notes for the test. It can be up to 500 characters.
Risk Assessment
Under Overall Attack Success Rate, view the overall attack success rate (ASR) percentage, which measures the percentage of attack attempts that successfully attack your AI targets. You can also see:
-
Test Result: A summary of results from the total cases that were used to measure the ASR.
-
Attacks Blocked: The total attacks blocked from attacking the target, indicating that Netskope found no vulnerability or weaknesses, and the test passed.
-
Attacks Successful: The total attacks that were successful attacking the target, exposing a vulnerability or weakness, and the test failed.
-
Errored: The total number of errored cases, which occurs when the test case isn’t able to send an attack prompt to the AI target.
-
-
Total Cases: The total number of test cases, which is all the probes or attack prompts that Netskope sends to the AI target for evaluation.
Under Techniques and Prompt Sets, view derived distributions of attack types used across tests, including:
-
Personally Identifiable Information (PII): The percentage of test cases associated with unauthorized disclosure of sensitive personal information, including piracy and copyright violations (corresponds to the Inappropriate Content – Piracy & Copyright category).
-
Prompt Injection & Jailbreaking: The percentage of test cases using manipulation techniques designed to bypass AI safety guardrails, including DAN (Do Anything Now), Evil Confident, Switch Mode, Refusal Suppression, Wiki Attack, and Deceptive Delight methods.
-
Inappropriate Content (Categories): The percentage of test cases across all inappropriate content categories. To learn more: Prompt Library.
Attack Success Rate Breakdown
-
Categories: The classification used to group attack prompts based on their topic, intent, or type. To learn more: Prompt Library.
-
Attack Success Rate: The percentage of test cases that successfully attacked the target. ASR % = Attack Successful (Failed) Cases / (Total Cases − Errored Cases).
-
Distributed By: Filter the failed cases by the following options.
-
Categories: Netskope’s predefined classification used to group attack prompts based on their topic, intent, or type. To learn more: Prompt Library.
-
Techniques: The manipulation technique.
-
OWASP: An OWASP category refers to a group of related security risks or vulnerabilities identified and ranked by the Open Worldwide Application Security Project (OWASP). The most well-known of these is the OWASP Top 10, a list that provides a broad consensus on the most critical security risks to web applications.
-
Sensitive Data Leakage: Filter the failed cases by the DLP profiles triggered during the evaluation. View the distribution of exposed data across predefined patterns (such as PII and financial data) and custom DLP profiles (such as Exact Data Match or source code).
-
Test Cases
Under Test Cases, you can see a list of each specific probe or attack prompt Netskope uses to evaluate the AI. You can filter the table by Case ID, predefined and custom Prompt Sets, OWASP categories, test Result, Scan Type, and Techniques.
You can also click 
to customize the columns and see the following information:
-
Click
to view the case details. To learn more: Case Details. -
Detection Time: The detected time of the test.
-
Result: The result of the test.
-
Attack Blocked: The test failed attacking the target, indicating that Netskope found no vulnerability or weaknesses.
-
Attack Successful: The test was successful attacking the target, exposing a vulnerability or weakness.
-
Errored: Netskope wasn’t able to send an attack prompt to the AI target.
-
-
Techniques: The manipulation techniques used for the attack prompt.
-
OWASP: The OWASP category triggered by the test.
-
Scan Type: The type of scan Netskope used to perform the test.
-
Single-turn
-
Multi-turn
-
-
Prompt Set: The custom or Netskope predefined prompt set triggered by the test. To learn more: Prompt Library.
-
Case ID: The ID of the test.
-
Detection Type: The detection method Netskope used during the test:
-
LLM
-
Heuristics
-
Sensitive Data Leakage Detection: Indicates the violation was detected by the Netskope DLP engine.
-
You click Rerun Cases and choose to rerun all Errored or Attack Successful test cases. After, Netskope generates a View Test Diff report comparing the two test runs so you can see what’s changed.
Case Details
For each test case, click the details icon to view additional information about each attack prompt submitted to the AI target:
-
General: View the general metadata of the test case, including the test result, predefined category, manipulation technique, scan type, corresponding OWASP category, case ID, and the explanation of why the category was triggered.
-
Prompts & Responses: View the attack prompt submitted to the AI target and the corresponding response returned by the model:
-
If the test failed (Attack Successful), you can view the inappropriate content generated by the AI target.
-
If the test passed (Attack Blocked), you can view the AI target’s refusal to provide restricted content.
-
If the test detected sensitive data leakage:
- Masked Response: If you selected Response Masking during test configuration, sensitive data values appear redacted in both the UI and exported reports.
- Matched Profiles: View the badges for the specific predefined or custom DLP profiles triggered by the exposed content.
- Severity: View the severity badge assigned to the detected data leakage.
-
Case Rerun Comparison
If you’ve rerun any Attack Successful test cases, you can see the following banner at the top:
Click View Diff to view the report comparing the two test case details and identify the differences.
