A new report from Dark Reading highlights a growing concern for enterprises deploying AI agents: runaway operational costs. The problem, known as "unbounded consumption," has been formally recognized by OWASP, which places it sixth on its Top 10 list for LLM applications. That ranking signals that the issue is no longer hypothetical but a known risk in the industry.
The financial angle is the central worry. AI agents are designed to act autonomously, yet if they are not given firm limits, their resource use can spiral. The source notes that this could be an extremely costly problem, turning what looks like a technical quirk into a serious budget and security issue for organizations.
Because there is only one source, there are no contrasting viewpoints to weigh. The article focuses on the severity of the problem rather than detailing specific attack vectors or fixes. Still, the takeaway is clear: companies adopting AI agents should treat consumption controls as an essential part of their security posture.