Why Gradual Disempowerment Remains a Credible Existential Risk

The author defends the 2025 Gradual Disempowerment paper against a recent critique, arguing that the critics overlook the core claim that competitive AI alternatives could erode human influence across all societal functions. They contend that even well-aligned individual AI systems would not prevent this systemic shift, as economic incentives would drive companies to shape policy and culture toward replacement. The piece concludes that the critique largely misses the paper's central argument about the conjunctive nature of human flourishing versus the disjunctive paths to catastrophe.

The Gradual Disempowerment paper, published in 2025 with a dedicated companion website, frames existential risk as arising not from a single rogue AI but from systemic replacement of human roles across labor, governance, and culture. Its authors argue that economic incentives alone would push companies to reshape policy and public opinion, creating a feedback loop that accelerates human marginalization even when individual AI systems remain corrigible.

Max Harms's defense emphasizes that the original paper's concern centers on correlated misalignment of large-scale social systems, not merely isolated AI failures. He notes that human flourishing requires many conditions to align simultaneously, whereas catastrophe can emerge through numerous independent pathways, making gradual disempowerment a credible threat alongside more dramatic takeover scenarios.