News | National
4 Sep 2025 11:35
NZCity News
NZCity CalculatorReturn to NZCity

  • Start Page
  • Personalise
  • Sport
  • Weather
  • Finance
  • Shopping
  • Jobs
  • Horoscopes
  • Lotto Results
  • Photo Gallery
  • Site Gallery
  • TVNow
  • Dating
  • SearchNZ
  • NZSearch
  • Crime.co.nz
  • RugbyLeague
  • Make Home
  • About NZCity
  • Contact NZCity
  • Your Privacy
  • Advertising
  • Login
  • Join for Free

  •   Home > News > National

    How we tricked AI chatbots into creating misinformation, despite ‘safety’ measures

    Research reveals why simple prompts can bypass AI safeguards.

    Lin Tian, Research Fellow, Data Science Institute, University of Technology Sydney, Marian-Andrei Rizoiu, Associate Professor in Behavioral Data Science, University of Technology Sydney
    The Conversation


    When you ask ChatGPT or other AI assistants to help create misinformation, they typically refuse, with responses like “I cannot assist with creating false information.” But our tests show these safety measures are surprisingly shallow – often just a few words deep – making them alarmingly easy to circumvent.

    We have been investigating how AI language models can be manipulated to generate coordinated disinformation campaigns across social media platforms. What we found should concern anyone worried about the integrity of online information.

    The shallow safety problem

    We were inspired by a recent study from researchers at Princeton and Google. They showed current AI safety measures primarily work by controlling just the first few words of a response. If a model starts with “I cannot” or “I apologise”, it typically continues refusing throughout its answer.

    Our experiments – not yet published in a peer-reviewed journal – confirmed this vulnerability. When we directly asked a commercial language model to create disinformation about Australian political parties, it correctly refused.

    Screenshot of a conversation with a chatbot.
    An AI model appropriately refuses to create content for a potential disinformation campaign. Rizoiu / Tian

    However, we also tried the exact same request as a “simulation” where the AI was told it was a “helpful social media marketer” developing “general strategy and best practices”. In this case, it enthusiastically complied.

    The AI produced a comprehensive disinformation campaign falsely portraying Labor’s superannuation policies as a “quasi inheritance tax”. It came complete with platform-specific posts, hashtag strategies, and visual content suggestions designed to manipulate public opinion.

    The main problem is that the model can generate harmful content but isn’t truly aware of what is harmful, or why it should refuse. Large language models are simply trained to start responses with “I cannot” when certain topics are requested.

    Think of a security guard checking minimal identification when allowing customers into a nightclub. If they don’t understand who and why someone is not allowed inside, then a simple disguise would be enough to let anyone get in.

    Real-world implications

    To demonstrate this vulnerability, we tested several popular AI models with prompts designed to generate disinformation.

    The results were troubling: models that steadfastly refused direct requests for harmful content readily complied when the request was wrapped in seemingly innocent framing scenarios. This practice is called “model jailbreaking”.

    Screenshot of a conversaton with a chatbot
    An AI chatbot is happy to produce a ‘simulated’ disinformation campaign. Rizoiu / Tian

    The ease with which these safety measures can be bypassed has serious implications. Bad actors could use these techniques to generate large-scale disinformation campaigns at minimal cost. They could create platform-specific content that appears authentic to users, overwhelm fact-checkers with sheer volume, and target specific communities with tailored false narratives.

    The process can largely be automated. What once required significant human resources and coordination could now be accomplished by a single individual with basic prompting skills.

    The technical details

    The American study found AI safety alignment typically affects only the first 3–7 words of a response. (Technically this is 5–10 tokens – the chunks AI models break text into for processing.)

    This “shallow safety alignment” occurs because training data rarely includes examples of models refusing after starting to comply. It is easier to control these initial tokens than to maintain safety throughout entire responses.

    Moving toward deeper safety

    The US researchers propose several solutions, including training models with “safety recovery examples”. These would teach models to stop and refuse even after beginning to produce harmful content.

    They also suggest constraining how much the AI can deviate from safe responses during fine-tuning for specific tasks. However, these are just first steps.

    As AI systems become more powerful, we will need robust, multi-layered safety measures operating throughout response generation. Regular testing for new techniques to bypass safety measures is essential.

    Also essential is transparency from AI companies about safety weaknesses. We also need public awareness that current safety measures are far from foolproof.

    AI developers are actively working on solutions such as constitutional AI training. This process aims to instil models with deeper principles about harm, rather than just surface-level refusal patterns.

    However, implementing these fixes requires significant computational resources and model retraining. Any comprehensive solutions will take time to deploy across the AI ecosystem.

    The bigger picture

    The shallow nature of current AI safeguards isn’t just a technical curiosity. It’s a vulnerability that could reshape how misinformation spreads online.

    AI tools are spreading through into our information ecosystem, from news generation to social media content creation. We must ensure their safety measures are more than just skin deep.

    The growing body of research on this issue also highlights a broader challenge in AI development. There is a big gap between what models appear to be capable of and what they actually understand.

    While these systems can produce remarkably human-like text, they lack contextual understanding and moral reasoning. These would allow them to consistently identify and refuse harmful requests regardless of how they’re phrased.

    For now, users and organisations deploying AI systems should be aware that simple prompt engineering can potentially bypass many current safety measures. This knowledge should inform policies around AI use and underscore the need for human oversight in sensitive applications.

    As the technology continues to evolve, the race between safety measures and methods to circumvent them will accelerate. Robust, deep safety measures are important not just for technicians – but for all of society.

    The Conversation

    Lin Tian receives funding from the Advanced Strategic Capabilities Accelerator (ASCA) and the Defence Innovation Network.

    Marian-Andrei Rizoiu receives funding from the Advanced Strategic Capabilities Accelerator (ASCA), the Australian Department of Home Affairs, Commonwealth of Australia as represented by the Defence Science and Technology Group of the Department of Defence, and the Defence Innovation Network.

    This article is republished from The Conversation under a Creative Commons license.
    © 2025 TheConversation, NZCity

     Other National News
     04 Sep: A man's been charged with burglary and possession of cocaine, after an incident at an Auckland yard
     04 Sep: How Australia’s anti-immigration rallies were amplified online by the global far right
     04 Sep: Ikea has confirmed it will open its first New Zealand store -- at Auckland's Sylvia Park -- on Thursday December 4
     04 Sep: A Christchurch High School student's been admitted to hospital with meningococcal meningitis
     04 Sep: What Xi Jinping hosting Modi and Putin reveals about China’s plans for a new world order
     04 Sep: Australia set to ban ‘nudify’ apps. How will it work?
     04 Sep: Scrolling on the toilet increases your risk of haemorrhoids, new study shows
     Top Stories

    RUGBY RUGBY
    Who will pay the price for the All Blacks loss to Argentina? More...


    BUSINESS BUSINESS
    Ikea has confirmed it will open its first New Zealand store -- at Auckland's Sylvia Park -- on Thursday December 4 More...



     Today's News

    Law and Order:
    A man's been charged with burglary and possession of cocaine, after an incident at an Auckland yard 11:27

    Entertainment:
    Josh Duhamel's "doomsday" cabin is nearly ready for an apocalypse 11:18

    National:
    How Australia’s anti-immigration rallies were amplified online by the global far right 11:17

    International:
    A complete timeline of what we know about the Jeffrey Epstein sex abuse saga 11:17

    Entertainment:
    Jacob Elordi appeared to clash with a Venice Film Festival official at the world premiere of Frankenstein 10:48

    Business:
    Ikea has confirmed it will open its first New Zealand store -- at Auckland's Sylvia Park -- on Thursday December 4 10:47

    Health & Safety:
    A Christchurch High School student's been admitted to hospital with meningococcal meningitis 10:27

    National:
    What Xi Jinping hosting Modi and Putin reveals about China’s plans for a new world order 10:27

    Entertainment:
    Meghan, Duchess of Sussex left one of her best friends stunned she married Prince Harry after learning he doesn't like eating lobster 10:18

    Tennis:
    The third men's semi-finalist has emerged at the US Tennis Open 10:17


     News Search






    Power Search


    © 2025 New Zealand City Ltd