Should Machines Have Rights?

Tools, not persons

Joanna Bryson argued under the title “Robots Should Be Slaves” (2010) – meaning, as she explained, servants one owns, not people – that robots are artefacts we design, build and own, and that we should neither build them to be persons nor treat them as such. She later called the wording a mistake but kept the argument. Ascribing robots rights or feelings wastes care and resources that belong to humans and lets those who design and deploy machines hide behind them. Because we decide what robots are, the responsible choice is to make them clearly recognisable as tools, so that responsibility stays with people.

Status from relations

Mark Coeckelbergh (2010) and David Gunkel (Robot Rights, 2018) doubt the usual procedure of first settling what a being is – conscious, sentient, rational – and then deriving its moral status. We rarely know such inner properties for certain, even in other people, and in practice we grant status through the ways we relate to others. Drawing on Levinas, Gunkel asks whether the encounter with the other, not a checklist of properties, comes first. The question of robot rights should therefore be opened, not closed in advance – without implying that robots must receive the same rights as humans.

Caution under uncertainty

A report by Robert Long, Jeff Sebo and colleagues, among them David Chalmers and Jonathan Birch (“Taking AI Welfare Seriously”, 2024), argues that there is a realistic possibility that some AI systems will be conscious or robustly agentic in the near future, and that AI companies should therefore acknowledge the issue, assess their systems and prepare policies. Birch (The Edge of Sentience, 2024) proposes proportionate precautions for all “sentience candidates”, from insects to possible future AI, while arguing that the fluent behaviour of language models is poor evidence of sentience, because it can be “gamed”. Eric Schwitzgebel, who defends rights for humanlike AI, warns of a “full rights dilemma”: for systems of debatable personhood, either choice risks grave moral error, so we should avoid building them. Metzinger goes furthest: since we could create artificial suffering without recognising it, research that aims at or risks synthetic consciousness should be suspended worldwide until 2050.

Held by

• David Chalmersb. 1966
• Thomas Metzingerb. 1958

Legal persons?

Legal personhood is not the same as moral status: companies are legal persons without feeling anything. In February 2017 the European Parliament asked the Commission to examine whether robots could eventually be given a legal status of their own, so that the most sophisticated autonomous robots could be treated as “electronic persons” liable for damage they cause. In 2018 more than 150 experts in robotics, AI, law and ethics objected in an open letter: such a status would rest on an overestimation of what current robots can do, and none of the available legal models – natural person, legal entity, trust – would fit. An expert group of the Commission concluded in 2019 that autonomous systems need no legal personality; the AI Act of 2024 regulates AI systems by risk level and places duties on their providers and deployers – it gives the systems themselves no legal status.