ABSTRACT
Legal scholars and policymakers assume the ‘right to be forgotten’ requires data controllers to erase personal information. This chapter demonstrates that assumption collapses for superintelligent AI. Neural networks store information as distributed parameters rather than discrete records, making selective deletion mathematically impossible. Machine unlearning fails because superintelligent systems reconstruct deleted data through inference.
Post-Singularity law must regulate AI behavior rather than data possession through five frameworks: (1) functional forgetting prohibits information from influencing outputs even if retained internally; (2) inference prohibition restricts conclusions systems may draw from legitimate data; (3) output governance regulates expression rather than knowledge; (4) memory fiduciary obligations impose duties of confidentiality and loyalty; and (5) cognitive sovereignty recognizes that human autonomy requires limits on what others remember.
Without international co-governance, the asymmetry between superintelligent recall and human forgetting will fundamentally undermine human identity construction and social forgiveness.
Chang, Cheng-chi (Kirin), Superintelligent AI and the Right to Be Forgotten (January 12, 2026) in: Research Handbook on Law in an Age of Superintelligence (Woodrow Barfield and Marc Jonathan Blitz eds, Edward Elgar, forthcoming 2027).
Leave a Reply