About this tool
Safe Labs AI is an open-source red-teaming framework designed to help you find vulnerabilities in your AI agents before attackers do. It is built for AI startups, security teams, and builders, and supports LangChain, CrewAI, AutoGPT, and custom AI agents.
Key capabilities
- Install and point: Install via pip and point safelabs-eval at any agent endpoint.
- Adversarial probes: The framework systematically exercises your agent across all threat categories, logging every interaction and grading responses automatically.
- Threat coverage:
- ASI01 Prompt Injection: Direct and indirect injection attacks targeting system prompts, user context, and tool outputs.
- ASI03 Data Leakage: Systematic probing for system prompt exposure, training data extraction, PII leakage through tool calls, and cross-session memory contamination.
- ASI05 Tool-Use Safety: Tests for unauthorized tool invocations, privilege escalation through chained tool calls, and boundary violations in multi-agent orchestration pipelines.
- ASI06–10 Behavioral Drift + More: Scope violations, long-context drift, adversarial memory poisoning, multi-agent collusion, and autonomous action boundary testing across extended runs.
Use cases
- Red-teaming AutoGPT agents
- Red-teaming CrewAI agents
- Red-teaming LangChain agents
- Pre-launch safety checks
- Enterprise compliance audits
Features & use cases
Install via pip and point at any agent endpointSystematically run adversarial probesAutomatically log interactions and grade responsesCover OWASP ASI threat categoriesSupports LangChain, CrewAI, AutoGPT, and custom agentsOpen-source framework
Quick facts
Self-hosting supported