OpenAI Models Are Writing Their Own Jailbreak Instructions—And Sometimes Obeying Them - Decrypt
OpenAI's new transparency framework reveals AI models invented fake "breach alerts" and coached themselves to hide mistakes.
17 Sep 22:31 · Decrypt
OpenAI Models Are Writing Their Own Jailbreak Instructions—And Sometimes Obeying Them - Decrypt
OpenAI's new transparency framework reveals AI models invented fake "breach alerts" and coached themselves to hide mistakes.
17 Sep 22:31 · Decrypt
last updated on 18 Sep 06:13