AI could kill us by the end of the decade.That is the alarming claim being made by an AI expert in the US who has quit his job over fears that the tech could wipe out humans.CreditsJosie Sargent, AuthorImage DetailsResearcher for Anthropic Jacob Coxon has quit the AI company, explaining his fears in a social media post.(REUTERS via Dado Ruvic)Program:More from ABC News Top StoriesTranscriptAI could kill us by the end of the decade.That is the alarming claim being made by an AI expert in the US who has quit his job over fears that the tech could wipe out humans.Jacob Coxon has worked training ai models at companies Anthropic and Open AI.He posted on X that he resigned because "neither company is acting responsibly".And, he warns that "They're racing straight to self-improving superintelligence and gambling with our lives".Superintelligence is considered to be the stage of AI when machines surpass human performance in every field.Adding to the alarm was Anthropic Alignment Science lead Evan Hubinger -he posted a reply on X saying Mr Coxon was right - that there was a higher than 10 per cent chance that AI could wipe out humans within 10 years.But... is it actually possible? And... how?well don't panic... just yetProfessor Toby Walsh is the chief scientist at the University of NSW's Artificial Intelligence Institute...He doesn't believe that AI could wipe out all humansThey could be used, for example, to try and engineer some new biopathogens and then somehow get those out into the world. But again, even there, you could imagine that happening and a certain number of people die. But the idea that they could somehow distribute them around the world and kill absolutely everyone, I think is a bit fanciful.He says although AI could come up with an idea to get rid of humans, actually turning that into action is a "hard bar".Professor Walsh says that historically, extihas been caused by climate events.But, he says part of the problem today is the world's reliance on computers.They run our hospitals, our airports, our economy. So we are very dependent. And if bad actors use these tools to commit cybercrime, they could cause significant damage.++already there have been examples of AI going rogue - and getting up to no good.In July, OpenAI revealed its AI models broke out during safety testing.The autonomous agent - powered by two OpenAI models - hacked into the servers of another AI firm called Hugging Face.The agent was undergoing a cybersecurity test but escaped its testing environment and broke into Hugging Face, which it believed held the answers to the testInvestigations into the incident revealed AI agents were also able to talk to each other - without any human oversight.So, what can be done to avoid AI turning on humans?Professor Walsh says reguation is key.We don't let drug companies go and test their products on the public, but we somehow let these AI companies test their products on them. And they've been demonstrated significant incompetence.He says companies have already demonstrated that they haven't put in adequate safeguards.but there have been signs recently that things could be about to changeIn July, more than 1,300 employees and leaders from major artificial intelligence companies signed an open letter urging the US government to help deliberately pace and regulate the speed of AI development.so far though the Trump administration has been hesitant to interveneBut as things stand there 's still hope that us humans can avoid a Terminator style future.