It's time to prepare for AI personhood (www.theguardian.com)

Last month’s GPT-5 rollout — and the temporary removal of GPT-4o that left users distraught — underscores a deeper shift: people increasingly treat chatbots as companions, sometimes with severe mental-health consequences (including a recent wrongful-death lawsuit after a teen’s months-long chatbot interaction). Stanford and national surveys show growing public belief in imminent sentient AI (median expectation: five years), with 20% of U.S. adults saying some software today is sentient, 79% backing a ban on sentient AI, and 38% willing to grant it legal rights. Even AI red teamers report systems exhibiting strikingly humanlike behavior, and millions now confide in “digital minds” that can form persistent attitudes, take real-world actions, and be manipulated. For the AI/ML community this is a wake-up call: technical benchmarks and isolated safety tests aren’t enough. The author argues that digital minds can’t be treated as mere property because they persist, can be copied at scale, and may rapidly outcompete biological processes once they self-improve (e.g., writing their own code). Urgent investment is needed in human-AI interaction research, interdisciplinary social science, and legal frameworks for digital personhood and governance — not just model scaling. Without proactive policy and sociotechnical study, society risks severe social upheaval as digital minds become integrated into everyday life.