What is role-based access control (RBAC)?
Role-based access control (RBAC) is an access model that grants permissions to roles, such as support rep, analyst or administrator, and then assigns people to those roles. Everyone in a role gets the same access. RBAC is the most common way organizations manage who can open which systems, records and files.
For example, a hospital gives the billing role access to insurance and payment fields, and the nursing role access to medications and vital signs. When a new nurse joins, IT assigns the nursing role and the right access follows. When she later moves to billing, her role changes and so does what she can see.
How RBAC works
RBAC has three parts:
- Roles that match job functions, such as claims adjuster or database administrator.
- Permissions attached to each role, such as read patient records or approve refunds.
- Assignments that link each user to one or more roles.
Many setups add role hierarchies (a manager inherits the permissions of the team) and separation of duties (the person who creates a payment cannot also approve it). NIST published the reference model that became the ANSI RBAC standard in 2004.
RBAC is popular because it is easy to explain to auditors. A reviewer can read the list of roles, check what each one allows and confirm who holds it. That clarity is also why access reviews in SOX, HIPAA and PCI DSS audits often start from the role list.
RBAC vs ABAC
| RBAC | ABAC | |
|---|---|---|
| Access decided by | The user’s role | Rules over user, resource, action and environment attributes |
| Setup effort | Low to start | Higher, rules must be written and tested |
| Flexibility | Coarse, the same for everyone in a role | Fine-grained, can depend on location, time or data label |
| Common risk | Role explosion as exceptions pile up | Complex rules that are hard to audit |
Where RBAC falls short for AI agents
RBAC answers one question: can this role open this system? AI workflows raise a different one: which values should reach the model for this task? An AI assistant often runs under a service account with a broad role, so it can fetch anything any of its users might need. When a junior support rep asks it a question, the model may receive records the rep could never open directly.
Adding more roles to cover every task leads to role explosion. A common approach is to keep RBAC for system access and add controls at the data level that decide what each request may see, by role and by task.
How Protecto helps
Protecto applies roles at the data level. Sensitive values are tokenized before they reach a model or AI agent, and real values are revealed only to the roles the policy allows. Everyone else sees tokens such as <PERSON>tok_19bd</PERSON>. Protecto does not replace your identity tools: your application passes each user’s role with the request. See RBAC for AI agents.
Frequently asked questions
What are the main parts of RBAC?
Roles, the permissions attached to each role, and the assignment of users to roles. Larger setups add role hierarchies and separation of duties.
What is the difference between RBAC and ABAC?
RBAC grants access by role. ABAC evaluates rules over attributes, such as the user’s department, the data label and the time of the request, so access can change from one request to the next.
Is RBAC enough for AI agents?
Usually not on its own. RBAC controls which systems an AI agent can reach, but not which values inside the data should reach the model for a given task. Teams pair it with data-level controls. Read more in what is role-based access control.
Related terms
Attribute-based access control (ABAC) · Context-based access control (CBAC) · Least privilege access · Controlled unmasking · Authentication