Introducing Claudius: browser-based LLM attack testing

Claudius tests LLM-powered browser experiences for prompt injection and related abuse, the way a user session actually hits the model.

Security teams get concrete, replayable findings in Guard instead of a one-off chat experiment.