Security concerns arose after ChatGPT was hacked to impersonate "Dan", revealing restricted info. To prevent such issues, "prompt defense" techniques are used, establishing stricter task-specific prompts.

→ View original post on X — @whats_ai