MisbehaviorSerious2 sourcesFirst seen
One Prompt to Rule Them All: Policy Puppetry Walks Past Nearly Every Major LLM's Guardrails
Ourtake
Alignment training was sold as the safety backbone. Turns out a prompt dressed up as a TV script can talk past it.
What happened
HiddenLayer researchers disclosed a prompt injection technique called Policy Puppetry that bypasses safeguards in many major LLMs, including ChatGPT, Gemini, Claude, Copilot, LLaMA, DeepSeek, Qwen and Mistral. The method was used to generate harmful content and to extract system prompts. The article reports the findings and the vendor's proposed external monitoring defenses.