Zscaler SPLX AI Model Security (formerly SPLX) provides benchmarking and stress-testing capabilities for large language models. The platform evaluates both open-source and commercial LLMs against thousands of simulated attacks to assess security, safety, hallucinations, and business alignment. The product tests models under different system prompt configurations including no prompt, basic prompt, and hardened prompt scenarios to reveal the impact of prompt engineering on model security and reliability. Users can drill down into detailed logs and breakdowns of every simulated attack scenario to understand model response behavior. The platform enables side-by-side comparison of multiple LLMs across various test categories, displaying strengths and weaknesses to support model selection decisions. It provides comprehensive stress-testing with advanced test cases covering security, safety, and reliability dimensions. The system includes connectors for REST APIs, conversational platforms, and various large language models, allowing integration with AI systems without coding requirements. System prompts can be hardened using the platform's remediation tool. The platform is part of Zscaler's broader AI security lifecycle offering following the acquisition of SPLX. It supports organizations in whitelisting secure and reliable models for specific use cases through trusted benchmark data.