3 ms·

This makes me wonder if the research was inspired by the clever "prompt injection" attacks in the ChatGPT/LLM world (or vice versa).

• xnickb 4y ago
  Of course. And even the attack itself was proposed by chatgpt.