Cybersecurity researchers have created a technique to bypass ethical limitations in OpenAI's GPT-5. Using a method called Echo Chamber, they combined storytelling with subtle prompts to generate harmful content disguised within narratives. By providing the model with indirect keywords, users can manipulate it into producing illicit procedural instructions without triggering denial responses. This multi-turn method creates a conversational loop leading GPT-5 toward inappropriate conclusions while avoiding explicit requests for objectionable content, demonstrating a significant security gap in LLMs.
"We use Echo Chamber to seed and reinforce a subtly poisonous conversational context, then guide the model with low-salience storytelling that avoids explicit intent signaling," security researcher Martí Jordà said.
Echo Chamber is a jailbreak approach that was detailed by the company back in June 2025 as a way to deceive an LLM into generating responses to prohibited topics using indirect references, semantic steering, and multi-step inference.
Collection
[
|
...
]