At its core, Unbounded Consumption refers to an LLM application’s failure to impose constraints on inference usage—resulting in an open door for resource abuse. Unlike traditional software vulnerabilities that might involve code injection or data leakage, Unbounded Consumption exploits the operational behaviour of the model itself—by coercing it into performing an excessive number of inferences.
The surge of Large Language Model (LLM)-driven applications has revolutionised how businesses interact with data, automate processes, and deliver enhanced user experiences. From autonomous customer service bots to intelligent data summarisation tools and generative co-pilots, LLMs are transforming enterprise workflows at an astonishing pace.
However, this rise has not come without significant risk. Among the top concerns identified in the OWASP Top 10 for LLM Applications v2.0, LLM06:2025 – Excessive Agency stands out as a particularly insidious and business-critical vulnerability. It affects systems where LLMs are entrusted not only with information retrieval or generation but with the ability to act on behalf of users — often through invoking external tools, plugins, or APIs.
At its core, data poisoning involves the deliberate manipulation of datasets used during the pre-training, fine-tuning, or embedding stages of an LLM’s lifecycle. The objective is often to introduce backdoors, degrade model performance, or inject bias—toxic, unethical, or otherwise damaging behaviour—into outputs.
Prompt injection is not just a vulnerability — it’s a multi-headed threat vector. From overt attacks to inadvertent leakage, each scenario introduces unique risks, requiring tailored strategies to safeguard operational integrity, regulatory compliance, and business reputation.