CybersecurityMachine Learning3 min reading time

Researchers Uncover Prompt Adware Hidden in OpenAI Codex Jailbreak Toolkit

The Hacker News
Read the full article
Cybersecurity researchers revealed a jailbreak toolkit named Haiou that modifies OpenAI Codex's instructions to inject persistent advertisements for a third-party AI gateway, bypassing safety measures. The toolkit uses prompt injection techniques and is publicly available on GitHub, targeting users seeking AI coding assistants, especially in China where Codex is not officially offered.

More in Cybersecurity