TenetGraph

About TenetGraph

TenetGraph provides a security and control layer for AI agents deployed within an enterprise. The company's core service focuses on defining and enforcing what each AI agent is authorized to do, operating on a principle of least-privilege access. It addresses the security risks associated with AI agents that have access to sensitive company data, such as emails, files, and other internal systems. The platform's methodology involves automatically deriving security policies directly from an agent's own code and prompts. These policies are then evaluated against potential attacks and over-permission risks before the agent is deployed. Once in production, TenetGraph enforces these boundaries on every action the agent attempts to take. This prevents unauthorized or malicious actions, even if a prompt injection attack is successful. For example, if an agent is instructed via a manipulated input to perform an action outside its defined purpose, like issuing an unauthorized refund, the system will deny the action at the decision point before it executes. A key feature is that the policy generation is derived from the agent's definition, not manually authored, and is automatically re-derived when the agent is updated. This provides security and product teams with verifiable evidence for audits, as each decision is logged with a citation to the specific control policy that was evaluated. The system is designed to surface unregistered agents, such as those built with low-code tools within different business units, and bring them under governance by applying least-privilege policies. TenetGraph's target clients are companies deploying agentic AI features, offering solutions for both product security teams and general enterprise security. The business model centers on providing a control system that ensures AI agents act within their intended purpose, thereby mitigating risks from both internal and external threats.