SplxAI (now part of Zscaler) is a specialized AI security platform built for automated red teaming and runtime protection. It provides comprehensive tooling for offensive security professionals to identify, exploit, and remediate vulnerabilities within the AI lifecycle, including LLMs, assistants, and autonomous agents.
Key Features
- Automated AI Red Teaming: Scales continuous testing using a vast database of AI-specific attack patterns and custom datasets to surface vulnerabilities like prompt injection.
- AI Asset Management: Automatically discovers AI models, workflows, and MCP servers to build a complete AI Bill of Materials (AI-BOM) and map the enterprise risk surface.
- Runtime Threat Inspection: Monitors LLM logs in near real-time to detect jailbreaks, prompt injections, and malicious queries during live interactions.
- Dynamic Remediation: Converts red teaming results into actionable hardening steps, including system prompt modifications to significantly reduce the attack surface.
- Agentic Radar: Provides an open-source security scanner (SAST) specifically for agentic AI, surfacing tool-level vulnerabilities and workflow-specific attack patterns.
Use Cases
- Offensive LLM Testing: Red teams can automate the discovery of adversarial bypasses in LLM guardrails and system prompts.
- AI Vulnerability Assessment: Penetration testers can utilize the platform to evaluate commercial and open-source models for safety and security weaknesses.
- Continuous Security Monitoring: Security operations teams can inspect live agentic workflows to detect unauthorized data access and real-time prompt injection attempts.
