Governing MCP Tool Permissions With RBAC
Control which MCP tools an agent can call, not just which servers.
Linnea Voss-Hartmann
Staff Writer, AI Security & Compliance
Linnea Voss-Hartmann
Linnea spent seven years as a cybersecurity analyst specializing in identity and access management before pivoting to technology journalism in 2018. Her reporting focuses on the intersection of regulatory frameworks and emerging AI deployment patterns, with particular depth on enterprise compliance failures.
14 stories
MCP Server Security Risks in Multi-Tenant Enterprise Environments
MCP servers lack built-in security controls, creating cascading risks across enterprise tenants.
Linnea Voss-Hartmann
MCP Server Authentication and Authorization Patterns
OAuth 2.1 and resource-bound tokens replaced MCP's trust-by-proximity model in just twenty months.
Linnea Voss-Hartmann
Least Privilege Access Enforcement for Agentic Workflows
Least privilege for agents requires task-scoped roles and permissions that expire automatically.
Linnea Voss-Hartmann
SAML and OIDC Flows for Non-Human AI Agent Identities
SAML and OIDC weren't built for agents acting without humans.
Linnea Voss-Hartmann
AI Agent Identity Standards and the Emerging Agent Identity Layer
Agents need identity systems built for nondeterministic, cloneable software, not humans.
Linnea Voss-Hartmann
AI Risk Assessment Framework for Enterprise AI Programs
Enterprises must map AI risks across shadow tools, data exposure, and agentic attacks.
Linnea Voss-Hartmann
Shadow AI Risks Specific to Enterprise Data Environments
Employees bypassing security to leak corporate data through unmonitored AI tools.
Linnea Voss-Hartmann
Cost Allocation and Chargeback for AI Agent Traffic
Agents multiply costs across six layers, but no dashboard reveals which team owes what.
Linnea Voss-Hartmann
Governing Employee Use of Public LLM APIs
Companies need technical controls, not policies.
Linnea Voss-Hartmann
AI Impact on Cybersecurity Operations and Threat Detection
Attackers now break in faster than SOCs can respond.
Linnea Voss-Hartmann
Audit Logging Requirements for Agentic AI Systems
Agents need audit logs that capture reasoning, not just actions.
Linnea Voss-Hartmann
Secrets and Credential Leakage Risks in AI Coding Assistants
AI-generated code leaks secrets at roughly double the baseline rate.
Linnea Voss-Hartmann
Third-Party AI Model Vendor Risk Assessment
EU AI Act deadlines force enterprises to formalize vendor governance now.
Linnea Voss-Hartmann