Liability, regulation, and AI’s new false dichotomy

Substack · Watch/Read · (gathered 2026-09-23)

• Gary Marcus refutes the “false dichotomy” that AI companies should face either liability or regulation, arguing that both are essential for ensuring safety and accountability, drawing parallels to other powerful industries.
• He highlights that relying solely on liability is insufficient because litigation is too slow (potentially taking a decade), leaves significant gaps in current laws (e.g., copyright, wholesale misinformation, Section 230 applicability), and would primarily benefit lawyers rather than effectively preventing harm.
• Marcus, supported by Karen Kornbluh, points to the aviation industry as a model where safety is achieved through a multi-layered approach combining government standards, verification, incident reporting/investigation, and liability for both violations and harms.
• He warns that without the right regulation tailored for AI, liability alone is “pointless” and will not prevent “a lot of bad things” from happening, especially given that tech companies can afford very expensive lawyers to navigate extensive litigation.