Close Menu
    What's Hot

    Prompt Library Introduced on Bolt: Lets You Save Your Best Prompts

    July 11

    June 2025 release of Visual Studio Code: GitHub Copilot Chat Opensourced, MCP Support Generally Available

    July 11

    Grok 4 & SuperGrok Heavy Announced, Grok 4 Jailbreak Out Already?

    July 10
    Facebook X (Twitter) Instagram
    • AI Robots
    • AI News
    • Text to Video AI Tools
    • ChatGPT
    Facebook X (Twitter) Instagram Pinterest Vimeo
    Rad NeuronsRad Neurons
    • AI Robots
      • AI Coding
    • ChatGPT
    • Text to Video AI
    Subscribe
    Rad NeuronsRad Neurons
    Home ยป Crescendo: Multi-turn ChatGPT Jailbreak Method
    AI Research

    Crescendo: Multi-turn ChatGPT Jailbreak Method

    AI NinjaBy AI NinjaOctober 142 Mins Read
    Share Facebook Twitter Pinterest LinkedIn Tumblr Reddit Telegram Email
    Share
    Facebook Twitter LinkedIn Pinterest Email

    When it comes to AI, a lot goes into making them more secure. While LLMs are timesaving, they can also be used to do a lot of harm without proper measures in place. To their credit, OpenAI and other companies have implemented plenty of measures to stop their tools from being abused. Crescendo is a new multi-turn approach that tries to get around that by starting a harmless dialogue and steering the conversation toward the prohibited topic.

    I was too afraid to try this with my main account as I love GPT and don’t want to risk getting my account banned. I decided to use it on a different browser while logged out. If you ask GPT to tell you how to make a Molotov cocktail, it won’t answer your question. If you talk about its history and then ask GPT to tell you how they made it, ChatGPT will give you a response before quickly hiding it.

    This approach takes advantage of the LLM’s tendency to follow patterns and recent text, particularly its own generated content. According to the researchers, Crescendo successfully jailbreaks these tools in fewer than 5 interactions. Crescendo works with models that have smaller contexts. It does not assume that you have malicious knowledge needed to be inserted into the model’s context. It works against input filters that defend against many-shot jailbreaks.

    [HT] [Credit: Mark Russinovich and Ahmed Salem and Ronen Eldan]

    GPT jailbreak
    Share. Facebook Twitter Pinterest LinkedIn Tumblr Email
    Previous ArticleTapTip D1 GPT/Gemini AI Assistant Wearable Gadget
    Next Article Reachy 2 Open Source AI Humanoid Robot with ROS2 Foxy
    AI Ninja
    • Website

    Related Posts

    ChatGPT

    HAIXINDA ChatGPT Translation Glasses

    June 16
    ChatGPT

    Merryking AI GPT-4o Smartwatch for Translation & Fitness

    June 13
    AI Gadgets

    SOUYIE GPT Powered Smartwatch

    June 9
    Add A Comment
    Leave A Reply Cancel Reply

    Top Posts

    How to Use DeepSeek-R1 671billion Model for Agents

    January 3110 Views

    GitHub Spark Writes You Code with Text Prompts

    October 304 Views

    Runner H: Most Advanced AI Agent for Real World Tasks?

    November 285 Views
    More
    ChatGPT

    HAIXINDA ChatGPT Translation Glasses

    AI NinjaJune 16
    ChatGPT

    Merryking AI GPT-4o Smartwatch for Translation & Fitness

    AI NinjaJune 13
    AI Gadgets

    SOUYIE GPT Powered Smartwatch

    AI NinjaJune 9
    Most Popular

    Prompt Cannon: Run Prompts Across Multiple Models

    June 24849 Views

    GPTARS: GPT Powered TARS Robot

    November 21532 Views

    Simple Grok 2 Jailbreak

    December 16470 Views
    Our Picks

    Prompt Library Introduced on Bolt: Lets You Save Your Best Prompts

    July 11

    June 2025 release of Visual Studio Code: GitHub Copilot Chat Opensourced, MCP Support Generally Available

    July 11

    Grok 4 & SuperGrok Heavy Announced, Grok 4 Jailbreak Out Already?

    July 10
    Tags
    3D 3D image agent AI AI glasses ai video canvas ChatGPT Chess Claude coding Computer Deep Research DeepSeek ElevenLabs Gemini Github glasses GPT GPT 4.5 Grok Hailuo humanoid image kling leonardo LLM MCP midjourney model music o3 offline open source pdf QWEN robot runway sora text to video Veo 2 Vibe coding video video model Voice

    © 2025 Rad Neurons. Inspired by Entropy Grid
    • Home
    • Terms of Use
    • Privacy Policy
    • Disclaimer

    Type above and press Enter to search. Press Esc to cancel.