Engineering safety
into AI.
U-Rec is an AI security company developing technology to evaluate and protect AI agents and robotics. Built on our research and development, we offer integrated support across custom AI development, security consulting, and training.
Book a free consultation
Introducing U-Rec
An era of AI that decides and acts.
Prompt injection, adversarial attacks, and unexpected autonomous behavior: as AI adoption grows, new risks emerge. U-Rec researches and develops technology to detect and defend against these threats, applying that expertise to AI system development and deployment support.
- Custom AI DevelopmentAI systems designed and developed with security built in.
- Vigil ↗Our runtime safety layer for AI agents and robotics.
- AI Security ConsultingFrom assessing AI use to establishing guidelines and operational processes.
- AI TrainingPractical training in the safe use of generative AI and AI coding tools.
Protecting AI in operation.
Vigil is a runtime safety layer for AI agents and robotics. It continuously safeguards AI in the field by detecting adversarial attacks, validating actions, and recording audit trails.
Explore Vigil ↗Insights
584 Incident Disclosures, Over Half Linked to Service Providers: Staying Ready Even Without a Direct Attack
A private database of public disclosures in Japan from June to October 2026 contains 584 information-leakage and security incidents. Over half spread from incidents at contractors, cloud platforms, and other services organizations use. If you suddenly learn of a possible breach, what should you stop, whom should you tell, and what should you fix first? We examine initial response, customer communication, statutory reporting, and prevention from management and operational perspectives.Approving Generative AI Use: How Much Do You Need to Check?
We'd like to use this AI at tomorrow's meeting. How should IT respond? Follow a meeting-notes example through three opening questions and copy a reusable decision note.Controlling AI Agent Permissions: NVIDIA's Safety Platform and Its Application to Robotics
A look at NVIDIA's Open Agent Safety Platform and runtime permission controls for AI agents. How OpenShell and Sentry differ, why API operations need limits, and what robots need for action validation and fault handling.