Common questions.
What is Inference-Time Attack?
An inference-time attack targets an AI agent during its operational phase, manipulating inputs, injecting content into tool outputs, or exploiting model weaknesses to produce attacker-controlled results.
How does Inference-Time Attack work?
Unlike training-time attacks, inference-time attacks can be carried out by anyone with access to the agent's input channels.
Which terms are related to Inference-Time Attack?
Closely related concepts include Agent Hijacking, Agent Retirement, Trust Chain, Unique Agent Identity. Each is defined in the Prefactor glossary.