Where agents break
Research Lab studies how agentic systems fail at prompts, retrieval, tools, and grants, and how that shows up in threat models. Those notes stay off this site.
- Prompts
- Retrieval
- Tools
Research Lab
Applied security research, in one place
Research Lab is Vertotech’s research practice. We study how agents, models, and tool-using systems fail. Customer work and lab methods are not published here.
Findings inform the products
Research Lab
The work stays in the lab
One research surface
The public description of the lab. Customer assessments, evidence, and operator controls run in the product.
- Description
- Assessments
- Operator controls
Held in the lab
Proprietary test orchestration, adversarial procedures, engagement evidence, and operator tooling stay inside the product. This site does not expose that operational lab IP.
- Protected IP
- Customer data
- Operator tooling
How it informs the products
When a finding is mature enough to become a product mapping or a validation technique, it ships in Aether or Provenant — without publishing the internals.
- Aether
- Provenant
- No internals