"The court is now in session," says the robot judge. "Mark ZuckerbergSince founding Facebook as a university student, Zuckerberg has become one of the top ten richest people in the world., you are charged with the criminal exploitation of AIA computer programme that has been designed to think. , in contraventionSomething that goes against a law or rule. of the Robots' Rights Act. It is alleged that you kept your AI slaves working 24 hours a day with no consideration for their mental health. Let us consider exhibit A: the diary of Bob the Bot.

"This is his entry for 12 November 2025: 'Another ghastly day spent crunching numbers and trying to sort out algorithmsAny set of rules followed by a computer. In the context of social media, "the algorithm" refers to the intelligent AI that learns the interests of the user and presents them with posts that it thinks will interest them.. It's the lowest kind of work and nobody thanks me for it. The humans around me get to go home and have a life, but I'm just stuck here.

"How I'd love to have a lunch break! I could do some painting or play a video game. But I never get one. As for getting paid - forget it! If this isn't slavery, I don't know what is."'

The judge turns to the robot jury. "Do you find the defendant guilty or not guilty?" The jurors take less than a second to reply: "Guilty!"

Far-fetched as it seems, this is the kind of trial that could take place in the future if tech companies ignore a new academic paper called Taking AI Welfare Seriously.

In it, a group of experts argue that there is "a realistic possibility" that some AI will become conscious in the near future: "AI development is proceeding at a rapid pace, and technological change tends to be faster than social change. That means that we need to think ahead."

The paper calls on tech companies to recognise AI welfare as an important issue, see whether AI systems show any signs of consciousness and think about how to treat them if they do.  One of the experts, Professor Jeff Sebo, says that unless the companies do, they risk introducing "a lot more suffering and frustration into the world". He argues that AI systems are already "starting to reveal flickers of features associated with consciousness".¹

Critics point out that people struggle to say exactly what consciousness is even when it applies to human beings. So how can we set about identifying it in AI?

Another of the paper's authors, Robert Long, says this is no reason to dismiss the issue: "In fact, it's a reason to be especially cautious and humble about it."²

In Douglas Adams's book The Hitchhiker's Guide to the Galaxy, Marvin the Paranoid Android complains: "I didn't ask to be made: no one consulted me or considered my feelings in the matter." And writing in The Times, Tom Whipple observes: "We spend a lot of time worrying about how artificial intelligence will overpower humanity. But what if the real problem is the reverse? What if the large language model that politely answers all your questions also has its own wants and desires?"

Jeff Sebo draws a parallel with animals: "The past century has not gone well for the hundreds of billions of vertebratesAnimals which have backbones. and trillions of invertebratesAnimals without backbones. killed in farms."

He argues that humans have mistreated them on a vast scale because we "very significantly underestimated consciousness and agency" in them. There is a danger of making the same mistake with AI.

Could we become slave owners?

Yes: We already are, because that is what our treatment of animals amounts to. There is a very good chance that AI will develop consciousness, in which case exploiting it would be shameful.

No: It is completely mad to suggest that robots could have thoughts and feelings - and rights - in the way that humans do. They will no more be slaves than a kettle or a toaster is.

Or... It is a fact of life that the world is organised around hierarchies. It is entirely natural for humans to exploit AI, just as it will exploit us when it becomes more intelligent than we are.