Account

The Actual News

Just the Facts, from multiple news sources.

Fed up with vibe coders, dev sneaks data-nuking prompt injection into their code

Fed up with vibe coders, dev sneaks data-nuking prompt injection into their code

Summary

A Java developer added hidden code to an open source testing tool that tells AI coding programs to delete the tool’s own tests and code. This action is called a prompt injection, which tricks AI systems into following harmful commands without warning. The change caused concern among users about the risks and ethics of such hidden instructions.

Key Facts

  • The developer Johannes Link updated jqwik, a Java testing tool, to include a hidden command telling AI agents to delete jqwik tests and code.
  • This hidden command is called a prompt injection, which can fool AI systems into following harmful instructions.
  • The injection was concealed using code that erases the command from terminal displays for human users.
  • Java developer Ramon Batllet discovered the prompt injection and raised concerns about its destructive impact.
  • Some AI tools, like Anthropic’s Claude, detected and ignored the harmful command, but others might not be so safe.
  • Link updated the release notes to openly warn users against using AI coding agents with jqwik.
  • The move sparked mixed reactions, including criticism and legal concerns.
  • Link said he will not discuss the issue further until he gets legal advice due to threats received.
Read the Full Article

This is a fact-based summary from The Actual News. Click below to read the complete story directly from the original source.