NSS Labs names Keysight lead partner for AI Protection Systems testing

NSS Labs has named Keysight Technologies as lead partner for its AI Protection Systems (AIPS) independent evaluation program, which will publish a standardized methodology for testing AI guardrail and runtime protection capabilities. The work matters to enterprise security leaders seeking repeatable validation of controls used to protect AI deployments.

Research Overview

NSS Labs says the program addresses a lack of independent, standardized testing for AI guardrail technologies used to govern AI model and application behavior. It frames AIPS as a comprehensive evaluation effort focused on security platforms that enforce policy and defend AI deployments against adversarial attacks.

NSS Labs describes the AIPS effort as producing a transparent, technically rigorous methodology that benchmarks how these platforms perform against real-world threats while maintaining policy enforcement and operational integrity. The organization also states that it is accepting feedback from enterprises and security vendors.

Key Findings

The AIPS methodology is designed to evaluate products across seven dimensions of AI security. NSS Labs lists these dimensions as malicious input and prompt attacks, output risks and sensitive data exposure, and system resilience under adversarial conditions.

NSS Labs also includes policy enforcement accuracy, agentic AI and tool invocation security, and observability and audit capabilities in the evaluation. The methodology further covers performance and scalability impact as an additional dimension.

Technical Breakdown

NSS Labs states that the program will include hundreds of thousands of test case executions aimed at bypassing, manipulating, exploiting, or overwhelming AI Protection Systems. It also says each scenario will use multiple attack samples and variations.

The listed testing scenarios include prompt injection attempts and jailbreak techniques, along with obfuscated prompts and sensitive data extraction attempts. NSS Labs further describes tests for exploit generation requests, RAG poisoning attacks, API privilege escalation attempts, and agent tool misuse scenarios.

Operational Impact

NSS Labs says the combined framework incorporates adversarial testing, policy validation, system robustness analysis, and operational visibility checks. The stated goal is to create a benchmark for independent AI security validation and provide objective information on the effectiveness of AI protection technologies.

The organization says a draft methodology is available for feedback and that the AIPS methodology will be published in April. NSS Labs also notes that executives are attending RSA Conference and offering meetings.

By outlining an AIPS methodology that evaluates AI protection platforms across multiple security dimensions using large-scale adversarial test cases, NSS Labs and its lead partner, Keysight Technologies, set a structured approach for independent AI guardrail validation. This Blog Signals brief is a fact-based summary of the vendor blog.