Longtermism is a moral and policy framework arguing that protecting the long-term future, including future generations and the avoidance of existential risk, deserves significant weight in present decision-making. It was most influentially developed by William MacAskill, Toby Ord, Nick Bostrom, and the Future of Humanity Institute lineage. The framework recurs in AI policy debates over how to weigh near-term harms against long-term and existential risks, and it is distinct from related concepts such as AI Existential Risk.
Background
The framework's value-claims component holds that future generations matter morally, and its empirical-claims component holds that extinction risk from advanced technology, including AI, is real. It has roots in the effective-altruism movement (Effective Altruism) and the existential-risk research associated with the Future of Humanity Institute lineage and AI-safety institutions in the FHI, MIRI, and FLI tradition.
The framework's canonical statements span three widely cited books: Nick Bostrom's Superintelligence (2014), which argued that a transition to machine superintelligence could pose extinction-level risk; Toby Ord's The Precipice (2020), which estimated roughly a one-in-six chance that humanity suffers an existential catastrophe this century and placed unaligned artificial intelligence among the largest contributors (Source: https://en.wikipedia.org/wiki/The_Precipice_(book)); and William MacAskill's What We Owe the Future (2022), which made the popular case that positively influencing the long-run future is a key moral priority of the present (Source: https://forum.effectivealtruism.org/posts/xmWAjdMyMKWxCRaTJ/overview-of-longtermism-its-ideas-writings-individuals). The more demanding academic version, "the case for strong longtermism," was set out by Hilary Greaves and MacAskill, holding that influencing the far future is the most important feature of many present actions.
The Future of Humanity Institute, the Oxford research center that anchored much of this work under Bostrom, was closed by the University of Oxford on 16 April 2024 after nineteen years, with the Institute citing "increasing administrative headwinds within the Faculty of Philosophy"; its closure is often read as marking the migration of existential-risk research out of a single academic hub and into a broader ecosystem of organizations (Source: https://www.futureofhumanityinstitute.org/).
In AI policy
Longtermism is the philosophical framing that places ASL-4+ catastrophic-risk thresholds (Anthropic RSP), AI Existential Risk framings, and existential-AI-risk priorities in conversation with near-term concerns such as bias, displacement, and misinformation. Several positions in the May 2026 AI policy debate reference the framing.
Aggressive AI timelines, as advanced in Situational Awareness: The Decade Ahead and by Jack Clark in Import AI 455, are partly motivated by longtermist concerns: if recursive AI R&D by end-2028 is plausible, present action carries disproportionate weight. The framing of "AI as normal technology" (AI as Normal Technology) functions partly as a critique of longtermism, arguing that diffusion-paced AI development justifies a near-term-only policy focus. The exclusion of Stuart Russell's testimony from Musk v. Altman (and OpenAI / Microsoft / Brockman), in which existential-risk framing was not admitted, represents the legal-frame version of the longtermism-versus-normal-technology debate.
Debates and positions
Civil-rights advocates often argue that prioritizing speculative future harms over documented present harms such as bias, mental-health effects, and displacement is wrong on its merits. A separate dispute concerns whether longtermism is best understood as an ideology or as an analytical lens: because the framing combines a value-claims component (future generations matter) with an empirical-claims component (extinction risk from AI is real), critics often conflate the two.
A sustained academic critique, associated with the philosopher Émile P. Torres among others, argues that strong longtermism can license neglect of present suffering by treating it as small relative to the vast potential value of the future, and that the framework concentrates moral authority in a narrow set of institutions and donors (Source: https://umbau.hfg-karlsruhe.de/posts/philosophy-against-the-present-the-foundations-and-critique-of-longtermism). Longtermism's standing in policy circles was further damaged in 2022 by the collapse of FTX, whose founder had been a prominent funder of effective-altruism and longtermist causes, a reputational shock that defenders and critics continue to dispute the lasting significance of.
Relationships
- related: AI Existential Risk, AGI Timelines, Effective Altruism.
- related: William MacAskill, Toby Ord, Nick Bostrom, Future of Humanity Institute (FHI).
- related: AI as Normal Technology — the principal counter-framing.
Sources
Expanded 2026-06-20 from supporting web sources (FHI closure statement; effective-altruism-forum overview of longtermism; published critiques) during a gap-scan thin-anchor pass. Foundational treatments (Bostrom, Ord, MacAskill) are cited by title; a foundational sources/ page for any of these would further firm up the page.