Gravitee warns companies about AI agent complexity and oversight gaps

In short: Gravitee says businesses are running into problems because AI agents are starting to interact in messy, hard-to-track chains, and oversight is not keeping up.

What's going on

Companies are not just using one AI agent at a time. They are deploying many agents that can call APIs, which are like digital doorways that let software talk to other software.

Gravitee CEO Rory Blundell says the real risk is not a single agent acting on its own. The risk is what happens when dozens or hundreds of agents start handing work to each other. A task that used to touch one system can now pass through several agents before a person sees it, and each handoff is another decision point.

Blundell argues that this complexity grows faster than most teams expect. Adding more agents can create many more possible connections between them, like adding more roads between cities, not just more cities.

He says common failures include “permissions creep,” where an agent is given broad access early on because it is faster, and then that access is never tightened later. Another issue is unclear ownership, where no specific person is responsible for what happens in the middle of a long agent chain.

What to watch

Gravitee says companies will likely need stronger “governance,” meaning clear rules, tracking, and enforcement. That includes giving each agent its own identity and limited access, plus tools that show, in real time, what an agent triggered downstream and can block actions that break policy before they happen.

Source: Venturebeat