A denial-of-wallet attack is a form of resource abuse that exploits usage-based billing or quotas to increase an organization’s costs or consume its available AI capacity. Attackers may generate excessive requests or otherwise cause an AI application to process large amounts of input and output. The resulting token consumption can drive up charges, exhaust a quota, or trigger throttling that limits service.
This matters because AI availability can depend on both provider limits and spending controls. Defenses include monitoring token and request usage, setting budgets and rate limits, limiting unnecessary output, and detecting unusual usage patterns. Teams should also test how critical workflows behave when requests are throttled or a quota is exhausted, with fallback procedures and human oversight where needed.