Forcepoint Warns AI Agents Could Drive Up Cloud Costs

AI agents can consume excessive computing power, tokens, API calls and other resources when safeguards and limits are missing, potentially leaving enterprises with higher-than-expected cloud bills, according to data security company Forcepoint.

Forcepoint calls the problem “unbound consumption.” The company says complex agentic AI systems can turn a single, apparently simple user request into tens or hundreds of downstream operations. High computing and token use may be difficult to detect, and existing security protocols may not catch it because no attacker is required: a misconfigured automation or a long-running AI session can drive up costs accidentally.

Attackers could also exploit the risk. Forcepoint says malicious users who obtain enterprise credentials could generate large workloads and consume computing resources, leaving the organization with the bill.

Forcepoint recommends setting budgets at multiple levels, including for API keys, individual users and teams, and improving monitoring of how costs are attributed. It also calls for “agentic circuit breakers” to stop workloads and costs from compounding.

Security researcher Jyotika Singh wrote that security teams rarely monitor cloud billing dashboards, while finance teams rarely review prompt patterns or agent design. She urged enterprises to take the risk more seriously.