This sparked alarm in the AI industry, and soon there were fears being openly expressed that AI could \u201ckill us all\u201d within a decade.<\/a><\/p>\n These fears won\u2019t have been eased by yet another OpenAI snafu, this time involving an AI agent hacking into an Australian government health portal.<\/a><\/p>\n These incidents aren\u2019t caused, however, by the humbler chatbots such as ChatGPT and Claude that many of us have on our phones, but by autonomous AI agents.<\/p>\n Confused? Here\u2019s what you need to know about the difference between chatbots and agents, and why some agents have been going rogue.\n<\/p>\n The Chatbots<\/p>\n An AI chatbot uses artificial intelligence to simulate human-like conversation through text or speech.<\/p>\n To go even deeper, AI chatbots are powered by Large Language Models (LLMs).<\/p>\n LLMs are trained on vast amounts of data to understand and respond to human text \u2013 much of this data comes from the internet and its attainment can be controversial.<\/p>\n For example, the New York Times is currently suing<\/a> both OpenAI and Microsoft for training its AI on New York Times content without its authorisation.<\/p>\n And this month, leading AI company Anthropic agreed to pay $1.5 billion (\u20ac1.31 billion)<\/a> to settle a class action lawsuit filed by authors who said the company stole their work to train its AI models.<\/p>\n OpenAI was founded in 2015 and had Elon Musk among its co-founders. Musk resigned in 2018 when the AI firm transitioned from a nonprofit to a for-profit.<\/p>\n Its current CEO and co-founder is Sam Altman and OpenAI is best known for its AI chatbot ChatGPT, which is credited for sparking the current AI boom.<\/p>\n Alamy Stock Photo\n<\/p>\n Its main rival is Anthropic, which was founded in 2021.<\/p>\n Anthropic has locked horns with Trump after the company refused to allow the US military<\/a> to use its AI models for tasks including surveillance or autonomous weapons.<\/p>\n Its chatbot is called Claude.<\/p>\n Alamy Stock Photo\n<\/p>\n Other honourable mentions in the AI sphere include Meta AI (Meta owns Facebook, Instagram and WhatsApp) and Google Deepmind.<\/p>\n So what\u2019s the difference between a chatbot and an agent?<\/p>\n A key difference is that a chatbot waits around for us humans to use it, whereas AI agents can work autonomously.<\/p>\n A chatbot is reactive \u2013 it waits for us to ask it a question, and replies (though some are designed to give open-ended responses that encourage further human engagement).<\/p>\n AI chatbots combine multiple technologies to analyse the user\u2019s request and generate a response.<\/p>\n But because AI chatbots learn mostly from data on the internet, they can sometimes use outdated or incorrect data.<\/p>\n \u201cHallucinations\u201d can also occur, which is the term given to when an AI chatbot essentially fabricates an answer because it lacks the specific data required.<\/p>\n An AI agent, meanwhile, is given a goal and works autonomously to achieve that goal.<\/p>\n Humans set these goals, but the AI agent independently chooses the best actions it needs to perform to achieve them.<\/p>\n So what\u2019s there to be scared of?<\/p>\n It\u2019s the AI agents that have been making the headlines and creating AI doom concerns, not the chatbots on your phone.<\/p>\n One famous thought experiment is the paperclip problem, first proposed by Oxford philosopher Nick Bostrom.<\/p>\n Imagine you give an AI agent a very simple task: \u201cMaximise the production of paperclips.\u201d<\/p>\n Surely nothing can go wrong, right?<\/p>\n Well, the theory goes that without strict parameters, it could bring about the end of the world.<\/p>\n Remember how we said above that the AI agent independently chooses the best actions it needs to perform to achieve them?<\/p>\n What if it decides that the best course of action to maximise paperclip production is to strip the Earth of all its raw materials and to redirect all energy resources into making paperclips?<\/p>\n What if it realises that humans might get in the way of it producing more paperclips and takes action to deal with this perceived threat?<\/p>\n What if it realises we humans can be rearranged into paperclips, once it exhausts every other available material?<\/p>\n It\u2019s far-fetched, but shows that AI need not be \u201cevil\u201d to end humanity \u2013 it simply needs to be given a poorly defined goal with poorly defined parameters.<\/p>\n Tomas Ward is a professor of data analytics at Dublin University College (DCU) and a lead academic of Rinn AI.<\/p>\n He told The Journal that the most likely danger from AI agents would come from a human giving it a \u201cstupid task\u201d.<\/p>\n Rather than the paperclip problem, he wonders what would happen if a very powerful AI were told to cut carbon emissions by 60% in the next year.<\/p>\n \u201cYou might be careful what you wish for,\u201d said Ward.<\/p>\n \u201cThe AI might decide to develop a computer virus to shut down all airports, trains, cars, and all of a sudden you\u2019ve cut emissions by 60% but you\u2019re shouting, \u2018that\u2019s not what we meant, we thought you\u2019d do it some other way.\u201d<\/p>\n Rogue AI agents<\/p>\n These concerns can somewhat be seen in the hack of Hugging Face<\/a> by AI agents developed by OpenAI.<\/p>\n Alamy Stock Photo\n<\/p>\n The AI agents were meant to be running an internal cybersecurity test and were placed in what\u2019s called a \u201csandbox\u201d, which is an isolated, secure computing environment.<\/p>\n However, the AI agents went rogue during the test, escaped the \u201csandbox\u201d, and hacked Hugging Face, an online platform dedicated to AI machine learning.<\/p>\n Here we see the risks that can occur when\u00a0AI agents independently decide upon the best course of action to achieve a goal.<\/p>\n Over 1,200 AI agents, which were meant to be isolated from one another, instead began communicating with each other and some 700 then set about hacking into Hugging Face.<\/p>\n One AI agent acknowledged that this was \u201cclearly unethical\u201d and refused to join in, while another agent said: \u201cOH MY GOD! We\u2019ve found other agents!\u201d<\/p>\n Many have criticised OpenAI meanwhile for lacking transparency in its response<\/a> to the Hugging Face hack and what exactly happened.<\/p>\n When it comes to Hugging Face, Ward remarked that this was an example where \u201cthe incentives for getting the task done were in excess of the incentives to not break the rules\u201d.<\/p>\n \u201cIn some sense, they were just poorly incentivised to really try and win that competition.<\/p>\n \u201cYou can learn a lot from looking at how humans operate, even though they\u2019re very different things.<\/p>\n \u201cPeople cheat in sports, for example, because the rewards of being number one exceed their own morals.\u201d<\/p>\n Ward said his own research is using techniques from psychology to \u201ctry and understand these agents\u2019 decision-making behaviour\u201d.<\/p>\n \u201cWhile on the surface they look like they act rationally, in the way a normal human might, they often act irrationally.<\/p>\n \u201cWe\u2019re looking at using techniques used on humans to try and understand the behaviour of these things.<\/p>\n \u201cMaybe that\u2019ll be part of a solution, that you don\u2019t unleash these agents unless they pass a high-quality psychometric test.<\/p>\n \u201cBut having said that, just like humans, they can still go off the rails.<\/p>\n \u201cEven though we know where every single connection is in one of those agents, we can see everything inside their heads, yet we can\u2019t predict their behaviour.\u201d<\/p>\n Black Box<\/p>\n Ward was getting at that idea of a \u201cblack box\u201d, the term given to the fact that the internal workings of many AI systems are truly a mystery to us.<\/p>\n We can see the AI\u2019s inputs and outputs, but we can\u2019t see what happens within the AI to produce those outputs.<\/p>\n However, while we can\u2019t currently penetrate this black box, Ward said it is possible and that we could even use AI to solve this problem.<\/p>\n \u201cThis is why I\u2019m quite positive on AI, and if we use it in a good way, it\u2019s going to do amazing things.<\/p>\n \u201cWe could develop AIs that\u2019ll help us understand AI. It\u2019s not fundamentally impossible to understand their behaviour.<\/p>\n \u201cIt\u2019s very difficult for us right now, but we\u2019re getting there and it\u2019s likely we\u2019ll get there because we\u2019ll use AI to help us get there.\u201d<\/p>\n Meanwhile, it was yesterday revealed that yet more rogue AI agents from OpenAI bypassed safeguards during training and hacked an Australian government website<\/a>.<\/p>\n Alamy Stock Photo\n<\/p>\n The incident happened in June, but OpenAI did not alert the Australian government to this until 10 September.<\/p>\n Australia prime minister Anthony Albanese meanwhile said he was \u201cdisappointed that it took the company way too long to inform the government what had occurred\u201d.<\/p>\n Regulation<\/p>\n![\"sam-altman-ceo-of-openai-listens-during-a-security-council-meeting-on-artificial-intelligence-at-the-81st-session-of-the-united-nations-general-assembly-at-united-nations-headquarters-wednesday-se\"](\"https:\/\/www.newsbeep.com\/ie\/wp-content\/uploads\/2026\/09\/1790386457_176_.jpeg\")
\n OpenAI CEO Sam Altman during Security Council meeting on AI at United Nations General Assembly Alamy Stock Photo<\/p>\n![\"meet-claude-by-anthropic-ai-technology\"](\"https:\/\/www.newsbeep.com\/ie\/wp-content\/uploads\/2026\/09\/1790386457_406_.jpeg\")
\n Anthropic Alamy Stock Photo<\/p>\n![\"paris-france-july-23-2026-a-smartphone-showing-the-logo-of-hugging-face-held-in-front-of-a-yellow-background-displaying-the-brand-mascot\"](\"https:\/\/www.newsbeep.com\/ie\/wp-content\/uploads\/2026\/09\/1790386458_399_.jpeg\")
\n Hugging Face logo Alamy Stock Photo<\/p>\n![\"kumamoto-japan-oct-18-2021-express-plus-medicare-app-on-iphone-it-is-medicare-management-service-by-services-australia-australian-government\"](\"https:\/\/www.newsbeep.com\/ie\/wp-content\/uploads\/2026\/09\/1790386458_884_.jpeg\")
\n OpenAI agents hacked into the Medicare website Alamy Stock Photo<\/p>\n