AI Policy Wiki
Dashboard
⚲
AI Policy Wiki › People & Organizations

PauseAI

medium confidence · updated 2026-07-22

Global grassroots movement founded in 2023 calling for a pause on frontier AI development; organizes protests, national chapters, and a treaty proposal for an international AI Safety Agency.

Contents

PauseAI is a decentralized global political movement that advocates a moratorium on the development of frontier AI systems, particularly the training of the most powerful models. It was founded in response to concerns about existential and catastrophic AI risk.

Type: Global political advocacy movement Mission: Advocate for a moratorium on the development of frontier AI systems

Founding and organization

PauseAI was founded in Utrecht, Netherlands, in May 2023 by Joep Meindertsma, a Dutch software entrepreneur who put his job on hold over what he described as existential risks from artificial intelligence; the group's first public action was a protest outside Microsoft's Brussels lobbying office (Source: pauseai.info). It has since grown into a network of national organizations — including PauseAI US (led by Holly Elmore), PauseAI UK, and chapters in countries including Germany, France, Australia, Kenya, and Nigeria — coordinated by a global organization whose chief executive is Maxime Fournes, with a board chaired by Meindertsma that includes Otto Barten (Source: pauseai.info). The global organization is registered as the Dutch foundation Stichting PauseAI (Source: pauseai.info).

Overview

PauseAI organizes protests, political campaigns, and advocacy for a global pause on training the most powerful AI models. Its members believe frontier AI development poses civilizational-scale risks, including double-digit percentage chances of catastrophic outcomes. Tyler Johnston of the Midas Project characterized PauseAI activists as people who "believe that, for example, AI development has some double-digit percentage chance of killing the people they love" (Source: washingtonpost.com).

Protest history

The organization's protest campaigns, by its own account, include a November 2023 protest outside the inaugural AI Safety Summit at Bletchley Park; a February 2024 gathering outside OpenAI's San Francisco headquarters; coordinated protests across thirteen countries in May 2024 ahead of the AI Seoul Summit; a larger multi-country protest for the February 2025 French AI Action Summit; and, in June 2025, its largest protest to that date outside Google DeepMind's London office over what the group called broken safety commitments, alongside its first PauseCon training event (Source: pauseai.info). A PauseCon organizing conference is planned for London in September 2026, and as of mid-2026 the group's campaigns include opposition to federal preemption of US state AI regulation (Source: pauseai.info).

Positions

PauseAI advocates full cessation of frontier development rather than slowing it, adding guardrails, or imposing safety requirements, placing it at the extreme end of the AI race dynamics debate. Its position is distinct from that of mainstream AI safety organizations such as Anthropic, Apollo Research, and METR, which accept continued development under safety frameworks.

The movement's premises — existential risk and an uncontrollable fast takeoff — are the claims rejected in the AI as Normal Technology debate advanced by Narayanan and Kapoor.

Treaty proposal

PauseAI's standing policy proposal (April 2026 version) calls for a temporary, globally coordinated pause on training the most powerful general AI systems, implemented through a treaty whose essential parties are the United States and China. The proposal's central measure is an international AI Safety Agency (AISA), modeled on the International Atomic Energy Agency, that would approve major training runs and deployments, set oversight requirements, and verify compliance through mechanisms such as GPU tracking, data-center inspections, and chip-based reporting; signatory coalitions would hold veto powers over deployment of superhuman AI systems. Supplementary measures include banning training on copyrighted material and holding model creators liable for criminal acts committed with their systems. The proposal states that the treaty would generally not affect narrow AI systems such as diagnostic image-recognition tools (Source: pauseai.info).

Connection to the AI safety violence debate

In April 2026, following attacks on Sam Altman's home and a shooting at an Indianapolis councilman's home that was accompanied by a "No Data Centers" note, industry voices argued that PauseAI and other AI "doomer" groups bore rhetorical responsibility for escalating violence. Safety advocates rejected that framing, arguing that the movement's concerns are legitimate even if individual acts of violence are inexcusable.

Tyler Johnston condemned the shootings as "awful" while describing the broader context: "We are entering a period of genuinely high stakes. There will be more radicals on every side, and some of them will resort to increasingly hostile action" (Source: washingtonpost.com).

Sources

Relationships

  • related: Future of Life Institute (FLI), Center for AI Safety (CAIS) — other organizations in the AI-safety advocacy ecosystem, though with different strategies and positions.
  • related: Stop AI — harder-line offshoot; its co-founders Sam Kirchner and Guido Reichstadter were expelled from PauseAI over their preference for civil-disobedience tactics involving law-breaking, which PauseAI rejects (Source: fortune.com).
  • related: AI Backlash — PauseAI anchors the organized-activism wing of broader public opposition to AI development.
  • related: AI Existential Risk — the risk claims motivating the movement.
Categories: People & Organizations | global | nonprofit | safety
What links here (12)