COMPEL Glossary / least-privilege
Least Privilege
Least privilege is a foundational security principle requiring that AI agents receive access only to the minimum set of tools, data, and system permissions necessary to perform their defined function.
What this means in practice
An agent designed to answer customer questions about order status should have read access to the orders database but should not have access to employee records, financial systems, or the ability to modify data. Each tool access permission should specify allowed operations (read vs. write), data scope (which tables or fields), rate limits (maximum queries per minute), and temporal constraints (valid during business hours only). Least privilege prevents both accidental damage (an agent querying the wrong database) and security exploitation (a compromised agent accessing sensitive systems beyond its purpose). In the COMPEL Agent Governance layer, least privilege is enforced through tool access controls with formal approval required for any expansion.
Why it matters
An AI agent with unrestricted access to enterprise systems is an unmanaged risk. Least privilege prevents both accidental damage from agents querying wrong databases and security exploitation from compromised agents accessing systems beyond their purpose. As organizations deploy more autonomous AI agents, the principle of least privilege becomes the primary mechanism for containing the blast radius of agent errors or security breaches.
How COMPEL uses it
In the COMPEL Agent Governance layer, least privilege is enforced through tool access controls with formal approval required for any expansion of agent permissions. During Model, each agent's tool access is defined with allowed operations, data scope, rate limits, and temporal constraints. The Produce stage implements permission controls. The Evaluate stage reviews tool access logs to verify agents operate within minimum necessary permissions.
Related Terms
Other glossary terms mentioned in this entry's definition and context.