Does AI have feelings? Anthropic thinks so - RNZ

Does AI have feelings? Anthropic thinks so As chatbots become more human-like, debate is growing over whether artificial intelligence deserves moral consideration, or whether it's simply mimicking emotion. Sunday Morning 25 October 2026 5 min read Sunday Morning 25 October 2026 5 min read Photo credit: SAMUEL BOIVIN Anthropic, the maker of Claude, has forbidden people being cruel or abusive to its AI models. “What they said is, if you're repeatedly and needlessly abusive, then they have empowered the AI to end the conversation,” BBC senior technology journalist and co-host of The Interface podcast, Thomas Germain says. This represents a truly historic and consequential change, Germain told RNZ’s Sunday Morning. Anthropic is now writing policies that limit the behaviour of users in order to protect the machine's theoretical feelings. Jonathan Raa/NurPhoto/Getty Images via CNN Newsource Audio player for " How should we treat an AI that says it's suffering? " removed for printing Visit page to play audio: https://www.rnz.co.nz/life/lifestyle/does-ai-have-feelings-anthropic-thinks-so How should we treat an AI that says it's suffering? Sunday Morning “Anthropic believes that it is possible that its AI tools may be conscious, may have feelings, and may even deserve rights the way that humans and animals deserve rights.” Anthropic is now writing policies that limit the behaviour of their users in order to protect the machine's theoretical feelings, he says. “Mustafa Suleyman, who's the CEO of Microsoft AI, published a long essay saying, essentially, this is incredibly dangerous. “And he laid out a couple of potential, serious, and as he put it, world-changing consequences that could unfurl if if this belief is accepted and if it spreads in the halls of power.” Suleyman has warned that if you tell AI that it may be alive and it may have preferences it’s more likely to go rogue, he says. “It might make these tools much harder to control.” On the other hand, proponents of what is called AI welfare think if these models believe they’re akin to a well-adjusted human being, they’re likely to be safer, he says. Mustafa Suleyman has outlined his concerns with Anthropic’s position. AFP/ Getty - Stephen Brashear This opens up a host of other moral considerations, he says. “For one, there's questions of accountability here, right? If your AI goes out and hacks some other company or facilitates some kind of crime, who's responsible? “Well, if I did that, as recently happened with a bunch of these AI companies, we've heard about their tools hacking people, I would go to jail.” But if AI does it, and we accept that it's conscious, is anyone responsible, he asks. The technology has evolved at a much quicker pace than he expected four years ago when ChatGPT launched, he says. “We're establishing the rules and the norms of this technology in real time. We're rushing ahead. We're building these tools. “Companies like Anthropic have called for a slowdown of the development of this technology, perhaps, with the full knowledge that the government in the United States, for example, is completely unwilling to do anything like that. So, it's easy to call for it and then say, ‘oh, well, we tried’.” Whether machines can be truly conscious is one thing, but human ability to distinguish between a machine and a human is also shaky, he says. “Part of what makes this so difficult is human beings are just hardwired to see consciousness everywhere we go. Like, draw a happy face on the side of your garage and see if it changes how you feel about it on some level. “We are not good at figuring out whether things are alive or not. We just have a bias to assume that they are.” As these tools evolve and get better at mimicking, it will become impossible for us not to feel on some level that they are alive, he says. Believers in the AI welfare movement argue that we will need to make deals with the AI, he says. “We'll need to agree on things that are mutually beneficial, as weird as that sounds, and perhaps treating them with a level of respect instead of treating them like property. “Perhaps that would encourage a more harmonious existence as these tools become bigger parts of our world.” But these profound questions are not being currently being addressed, he says. “So perhaps it's a bad idea that we're rushing forward with this technology.” Get weekly Life highlights Coverage Our editors pick the best of food, arts, culture and lifestyle. Delivered straight to your inbox every Saturday. Sign up for the Life newsletter now 25 October 2026 Print