We work with organizations that develop models, AI infrastructure, and AI products. We run red team testing of models, agents, and guardrails in production environments.
§ 01What you're up against
The threats characteristic of your sector.
01
Prompt injection
Direct and indirect prompt injection attacks against your applications and the data sources they use.
02
Agent and tool abuse
When an AI system can take actions, we assess what happens if an attacker attempts to steer them.
03
Model and data poisoning
We examine training and retrieval pipelines for manipulation and leakage.