prompt injection

名词 n.

英文释义

名词 n.
  1. A method of manipulating an artificial intelligence system by embedding malicious instructions (prompts) within input data in order to trick the model into overriding its intended behaviour or safeguards. countable,uncountable
    — Yesterday, OpenAI announced an API for ChatGPT and posted an underlying format for the bot on GitHub, alluding to the issue of prompt injections.
  2. A method of causing an artificial intelligence to ignore its initial instructions (often ethical restrictions) by directly giving it a particular prompt; a jailbreak. countable,uncountable
    — Retomeli.io is a jobs board for remote workers, and the website runs a Twitter bot that spammed people who tweeted about remote working. The Twitter bot is explicitly labelled as being "OpenAI-driven", and within days of Goodside's proof-of-concept being published, thousands of users were throwing prompt injection attacks at the bot.

词形变化

prompt injections plural
0 次浏览 数据来源: Wiktionary