← Glossary Company
Adversa AI
A security research firm that specialises in breaking AI systems and publishing how it did it.
Adversa AI is a small research outfit working on what the industry calls red teaming for AI: deliberately attacking models and the products built around them to find out where they give way. Its published work tends to focus on the seam where a chatbot stops being a text box and starts doing things, browsing pages, running code, reading your files.
The firm has a habit of naming its findings, which is how techniques like Cryptographic Context Injection end up with a label the rest of the industry then adopts. It reports through the usual disclosure channels first, and publishes whether or not the vendor has fixed anything.
Mentioned in