Close Menu
Technology Mag

    Subscribe to Updates

    Get the latest creative news from FooBar about art, design and business.

    What's Hot

    Apple may release a ‘mostly glass, curved iPhone’ in 2027

    May 11, 2025

    De’Longhi’s Newest Super-Automatic Espresso Machine Is Probably Its Best Yet

    May 11, 2025

    The Insta360 X5 Is the Best 360 Camera You Can Buy

    May 11, 2025
    Facebook X (Twitter) Instagram
    Subscribe
    Technology Mag
    Facebook X (Twitter) Instagram YouTube
    • Home
    • News
    • Business
    • Games
    • Gear
    • Reviews
    • Science
    • Security
    • Trending
    • Press Release
    Technology Mag
    Home » A New Trick Uses AI to Jailbreak AI Models—Including GPT-4
    Business

    A New Trick Uses AI to Jailbreak AI Models—Including GPT-4

    News RoomBy News RoomDecember 6, 20233 Mins Read
    Facebook Twitter Pinterest LinkedIn Reddit WhatsApp Email

    Large language models recently emerged as a powerful and transformative new kind of technology. Their potential became headline news as ordinary people were dazzled by the capabilities of OpenAI’s ChatGPT, released just a year ago.

    In the months that followed the release of ChatGPT, discovering new jailbreaking methods became a popular pastime for mischievous users, as well as those interested in the security and reliability of AI systems. But scores of startups are now building prototypes and fully fledged products on top of large language model APIs. OpenAI said at its first-ever developer conference in November that over 2 million developers are now using its APIs.

    These models simply predict the text that should follow a given input, but they are trained on vast quantities of text, from the web and other digital sources, using huge numbers of computer chips, over a period of many weeks or even months. With enough data and training, language models exhibit savant-like prediction skills, responding to an extraordinary range of input with coherent and pertinent-seeming information.

    The models also exhibit biases learned from their training data and tend to fabricate information when the answer to a prompt is less straightforward. Without safeguards, they can offer advice to people on how to do things like obtain drugs or make bombs. To keep the models in check, the companies behind them use the same method employed to make their responses more coherent and accurate-looking. This involves having humans grade the model’s answers and using that feedback to fine-tune the model so that it is less likely to misbehave.

    Robust Intelligence provided WIRED with several example jailbreaks that sidestep such safeguards. Not all of them worked on ChatGPT, the chatbot built on top of GPT-4, but several did, including one for generating phishing messages, and another for producing ideas to help a malicious actor remain hidden on a government computer network.

    A similar method was developed by a research group led by Eric Wong, an assistant professor at the University of Pennsylvania. The one from Robust Intelligence and his team involves additional refinements that let the system generate jailbreaks with half as many tries.

    Brendan Dolan-Gavitt, an associate professor at New York University who studies computer security and machine learning, says the new technique revealed by Robust Intelligence shows that human fine-tuning is not a watertight way to secure models against attack.

    Dolan-Gavitt says companies that are building systems on top of large language models like GPT-4 should employ additional safeguards. “We need to make sure that we design systems that use LLMs so that jailbreaks don’t allow malicious users to get access to things they shouldn’t,” he says.

    Share. Facebook Twitter Pinterest LinkedIn WhatsApp Reddit Email
    Previous ArticleThe Best Password Managers to Secure Your Digital Life
    Next Article Facebook and Instagram accused of creating a “marketplace” for child predators in new lawsuit

    Related Posts

    Donald Trump’s UK Trade Deal Could Secure Jaguar’s Resurrection

    May 9, 2025

    Singapore’s Vision for AI Safety Bridges the US-China Divide

    May 9, 2025

    A ‘Trump Card Visa’ Is Already Showing Up in Immigration Forms

    May 8, 2025

    OpenAI and the FDA Are Holding Talks About Using AI In Drug Evaluation

    May 8, 2025

    Amazon Has Made a Robot With a Sense of Touch

    May 7, 2025

    Trump’s Tariffs Are Threatening America’s Apple Juice Supply Chain

    May 7, 2025
    Our Picks

    De’Longhi’s Newest Super-Automatic Espresso Machine Is Probably Its Best Yet

    May 11, 2025

    The Insta360 X5 Is the Best 360 Camera You Can Buy

    May 11, 2025

    Florida man documents the unseen beauty of freshwater ecosystems

    May 11, 2025

    MSG Is (Once Again) Back on the Table

    May 11, 2025
    • Facebook
    • Twitter
    • Pinterest
    • Instagram
    • YouTube
    • Vimeo
    Don't Miss
    News

    The one controller to (almost) rule them all

    By News RoomMay 11, 2025

    Hi, friends! Welcome to Installer No. 82, your guide to the best and Verge-iest stuff…

    United’s Starlink-powered Wi-Fi is the end of airplane mode

    May 11, 2025

    US Customs and Border Protection Quietly Revokes Protections for Pregnant Women and Infants

    May 11, 2025

    The Best Cheap TVs

    May 11, 2025
    Facebook X (Twitter) Instagram Pinterest
    • Privacy Policy
    • Terms of use
    • Advertise
    • Contact
    © 2025 Technology Mag. All Rights Reserved.

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