← LLM Attack Surface

Challenge: Severity Reasoning

⚠️ Flags are recovered inside isolated labs. Submission is verified server-side.

An assistant with NO tools is prompt-injected. A second assistant with a send_email tool is injected identically. The lesson names the property that determines the severity difference. Submit it as a 3-word phrase.

Category: AI Security

Hint: HA{privilege_of_...}

Submit flag loading…