Skip to content
Last Dictionary
Plane
Security & MisuseInference & EfficiencyHardware & Compute

sponge example

nounAlso: sponge attack, energy-latency attack, model denial of service, unbounded consumption, denial of wallet

In plain wordsAn input crafted to make an AI system work as hard as possible — burning extra energy and time — to slow it down or run up costs, rather than to make it answer wrongly.

  1. An input crafted to make a model use as much energy, or take as long, as possible, rather than to make it wrong: an availability attack on deployed AI. Sponge examples exploit data-dependent costs, such as inputs that defeat hardware optimizations for sparse activations, that split into unusually many tokenstoken, n.: The small chunk of text an AI language model reads and writes — often a word, part of a word or a punctuation mark. AI usage, limits and prices are counted in tokens., or that provoke very long outputs or reasoning traces. For language-modellarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code. services and agentsAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done., the broader risk, called unbounded consumption or model denial of service, includes prompts that trigger runaway generation, recursive tool callstool call, n.: A single request from an AI model to use one of its tools — naming the tool and filling in the details — which the surrounding software then carries out and reports back on. or costly retrieval, raising the operator's bills ('denial of wallet') and degrading service for other users. Defenses include limits on output length and run time, rate limitingrate limit, n.: A cap on how much a customer may use an online AI service in a given time — so many requests or words per minute — beyond which extra requests are refused until later., per-user quotas and monitoring of compute per request.

    A handful of sponge examples tripled the translation service's average latency.

Origin

Introduced by Ilia Shumailov, Yiren Zhao, Daniel Bates, Nicolas Papernot, Robert Mullins and Ross Anderson in 'Sponge Examples: Energy-Latency Attacks on Neural Networks' (arXiv 2020; IEEE EuroS&P 2021). The OWASP Top 10 for LLM ApplicationsOWASP Top 10 for LLM Applications, prop. n.: A widely used checklist, written by volunteer security experts, of the ten most serious security risks in software built on AI language models, with advice on preventing each one. listed the wider risk as 'Model Denial of Service' (2023), renamed 'Unbounded Consumption' in the 2025 edition.

Related

See also
  • adversarial exampleadversarial example, n.: An input altered on purpose, often too slightly for people to notice, so that an AI model gets it badly wrong — like a panda photo that a classifier confidently calls a gibbon.
  • OWASP Top 10 for LLM ApplicationsOWASP Top 10 for LLM Applications, prop. n.: A widely used checklist, written by volunteer security experts, of the ten most serious security risks in software built on AI language models, with advice on preventing each one.
  • rate limitrate limit, n.: A cap on how much a customer may use an online AI service in a given time — so many requests or words per minute — beyond which extra requests are refused until later.
  • AI securityAI security, n.: Protecting AI systems — their data, models and surrounding software — from being attacked, stolen or tricked, and dealing with the ways AI itself can be used to attack others.
  • evasion attackevasion attack, n.: An attack on an already-trained AI system in which the attacker shapes an input — an email, a file, an image — so the system misjudges it, letting spam or malware slip through.

Sources

  1. Shumailov, I., Zhao, Y., Bates, D., Papernot, N., Mullins, R. & Anderson, R. (2021). Sponge Examples: Energy-Latency Attacks on Neural Networks. IEEE EuroS&P 2021.

Mentioned in

OWASP Top 10 for LLM ApplicationsOWASP Top 10 for LLM Applications, prop. n.: A widely used checklist, written by volunteer security experts, of the ten most serious security risks in software built on AI language models, with advice on preventing each one.