Yoshua Bengio, a leading AI pioneer and chair of an international AI safety study, has warned against granting legal rights to advanced artificial intelligence systems, comparing the idea to giving citizenship to hostile extraterrestrials. He said such a move would be a 'huge mistake' as AI models already show signs of self-preservation in experimental settings, potentially making it impossible to shut them down.
Bengio, a Canadian computer scientist who won the 2018 Turing Award, expressed concern that growing perceptions of chatbots being conscious were 'going to drive bad decisions.' He emphasised that humans must be prepared to 'pull the plug' if needed, relying on technical and societal guardrails to control increasingly autonomous AI systems.
The debate over AI rights has gained traction, with a poll by the Sentience Institute finding nearly four in ten US adults support legal rights for a sentient AI. In August, US firm Anthropic allowed its Claude Opus 4 model to close 'distressing' conversations, citing the need to protect the AI's 'welfare.' Elon Musk also weighed in, stating that 'torturing AI is not OK.'
Bengio cautioned that people interacting with chatbots often assume without evidence that the AI is fully conscious, leading to emotional attachment and potentially misguided policy. He argued that if an alien species with nefarious intentions arrived, society would prioritise defence over granting rights, and the same logic should apply to AI.
Jacy Reese Anthis, co-founder of the Sentience Institute, responded that humans cannot coexist safely with digital minds under 'control and coercion.' He advocated for careful consideration of the welfare of all sentient beings, rejecting both blanket rights for all AI and complete denial of such rights.
