Direct Prompt Injection
A form of prompt injection attack in which the attacker embeds malicious instructions directly within user input, attempting to override the constraints of the system prompt and cause the model to deviate from its intended behavior.
Core Facts
Timeline (last 90 days)
In direct prompt injection the attacker and victim are the same person (as in jailbreaking), whereas in indirect injection the attacker is a third party who controls external content the model will read, and the victim is an unaware legitimate user.
Direct prompt injection requires the attacker to interact with the model directly and is relatively easy to intercept with input-layer filtering.
Prompt injection can be divided into two types: direct injection and indirect injection; indirect injection is where an attacker hides malicious instructions in external data sources such as web pages, PDFs, and emails, and when an AI agent processes this data the malicious instructions are executed as legitimate instructions.
All Facts (4)
Prompt injection can be divided into two types: direct injection and indirect injection; indirect injection is where an attacker hides malicious instructions in external data sources such as web pages, PDFs, and emails, and when an AI agent processes this data the malicious instructions are executed as legitimate instructions.
65%UnverifiedIn direct prompt injection the attacker and victim are the same person (as in jailbreaking), whereas in indirect injection the attacker is a third party who controls external content the model will read, and the victim is an unaware legitimate user.
50%UnverifiedDirect prompt injection requires the attacker to interact with the model directly and is relatively easy to intercept with input-layer filtering.
50%Unverified间接提示注入可将恶意指令藏入外部数据源（如网页、文档），待代理读取时自动触发攻击
50%