AI systems increasingly access source code, files, credentials, APIs, cloud environments, and internal data. This makes AI part of the organization’s security and trust architecture.
Cyfinoid researches how AI can be deployed and used while retaining control over data, infrastructure, permissions, and model actions.
Research Focus
Private AI Infrastructure
We examine local and cloud-hosted AI deployments, model gateways, request filtering, data leakage controls, logging, hardware planning, and the trade-offs between self-hosted and externally hosted models.
AI Agent Security
Our work covers coding agents, tool calling, MCP servers, credential exposure, project isolation, network access, persistent agent state, and software supply chain risks introduced by AI-assisted development.
AI Supply Chain
AI applications depend on models, libraries, cloud services, datasets, infrastructure, and hardware. We explore practical methods to identify, document, and assess these components.
Community Contributions
AI BOM GeneratorAnalyze public GitHub repositories for AI libraries, models, infrastructure, hardware requirements, and governance indicators, then generate experimental AI BOM output
AIDC: AI Development ContainersRun AI coding agents inside project-specific containers with restricted host access, persistent agent state, built-in security scanners, and optional network controls.
Council of AI bots🧠Simulate a council of AI personas. Multi-perspective reasoning — now in your browser. AI Bot Council is a browser-based experiment in collaborative AI…No projects match the selected filters.
Blogs
Why This Matters
The security boundary of an AI system includes the model, its data, tools, identity, infrastructure, and the workflows that act on its output.
Secure AI adoption requires visibility into what the system can access, where information is sent, and which actions it can perform.
Cyfinoid turns this research into open-source tools, technical guidance, training, demonstrations, and security consulting.