AWS roles are multiplying faster than teams can keep track. One month you have a clear access model; the next, your IAM console looks like a junk drawer of overlapping policies, temporary exceptions, and one-off privileges. This is the large-scale role explosion problem, and it is one of the quietest, most dangerous scaling risks in cloud infrastructure.
First, it slows you down. Every new role, policy, and trust relationship makes it harder to review permissions or predict the blast radius of a misconfiguration. Then, it multiplies risk. Redundant or unused roles often have more privileges than intended, and attackers know how to find them.
The root cause hides in plain sight. Multiple teams, moving fast, create one-off AWS roles to meet immediate needs. Short-term expedience slowly builds a sprawling inventory of access objects that no one owns but everyone assumes are safe. Cross-account trust policies, inline permissions, service-linked roles, and external identity providers make the surface area even bigger.
Cleaning it up is not simple. You can’t just delete roles without breaking workloads, and you can’t freeze changes without grinding delivery to a halt. Large organizations try audits, spreadsheet inventories, or custom scripts, but these solutions rarely keep pace with actual change. Even advanced tooling from cloud providers can leave blind spots, especially when roles interact across accounts or connect to third-party services.