Adversarial AI
Find where prompt injection and other adversarial inputs can redirect an application or agent.
Promptention RedPeople and research
Join our talent pool.
We investigate attacks against AI applications and downloaded model files, then turn that work into Guard, Red, and Model Clearance. If your work overlaps, we would like to see it.
Share your workWork behind the products
These areas shape what we research, test, and build. Tell us where your work connects.
AI application security
Work on the runtime boundaries around prompts, retrieval, tool use, and model output.
Promptention GuardModel and artifact security
Examine model formats, loading behavior, dependencies, and executable structures inside files.
Promptention Model ClearanceEvaluation and measurement
Design repeatable tests and investigate false alarms so product findings stay useful.
Across our productsSecurity product engineering
Turn security research into dependable inspection and analysis workflows.
Across our productsTechnical communication
Explain attack paths, findings, and limitations so other teams can act on the work.
Research and productTell us what you've been working on.
A short note or one link to research, code, or writing is enough to start a conversation.
Write to us
We are exploring equity and commission based arrangements. Terms depend on the work.