{"slug": "is-a-pause-enforceable-new-paper-out", "title": "Is a pause enforceable? New paper out!", "summary": "A new paper by Raymond Koopmanschap and Otto Barten, titled 'How to Catch a GPU: A Taxonomy of Verification and Enforcement Mechanisms for International AI Agreements,' argues that enforcing a pause on large AI training runs is possible if training runs are large enough, and introduces the concept of a 'breakdown point' in FLOPs or H100-equivalents below which regulation becomes unenforceable. The paper investigates eight off-switch proposals, catalogs policies per sub-problem, and provides tables with enforcement breakdown points and FLOP thresholds.", "body_md": "**Some say, a pause is possible**, it merely depends on political will. Others say, tech progress can never be stopped. Maybe the truth lies in the middle? Maybe we can 'catch all GPUs' (that is, effectively enforce a pause or other international agreement) as long as training run sizes are large enough? Maybe every government policy would therefore have a *breakdown point*, a number of FLOPs or H100-equivalents where the training run gets too small to enforce regulation?\n\nIn our new paper, [ How to Catch a GPU: A Taxonomy of Verification and Enforcement Mechanisms for International AI Agreements](https://arxiv.org/abs/2607.22619):\n\nIn addition, we investigate eight off switch/treaty verification proposals from the literature, and catalogue policies they propose. We think our tables with proposed policies, structured per sub-problem, are particularly useful for researchers:\n\nIn addition, we include a table with enforcement breakdown pionts per measure for the determining sub-problem *detect hidden capacity*:\n\nAnd finally, we include a table in the appendix with proposed FLOP thresholds per off switch proposal:\n\nWe hope this work will be useful!\n\nFull paper: [https://arxiv.org/abs/2607.22619](https://arxiv.org/abs/2607.22619)\n\nCite as (MLA): Koopmanschap, Raymond, and Otto Barten. \"How to Catch a GPU: A Taxonomy of Verification and Enforcement Mechanisms for International AI Agreements.\"\n\nEA Forum post: [https://forum.effectivealtruism.org/posts/9yxcgyAmgFDyEFb2W/is-a-pause-enforceable-new-paper-out](https://forum.effectivealtruism.org/posts/9yxcgyAmgFDyEFb2W/is-a-pause-enforceable-new-paper-out)", "url": "https://wpnews.pro/news/is-a-pause-enforceable-new-paper-out", "canonical_source": "https://www.lesswrong.com/posts/qygnNCM2FAA76Z9T4/is-a-pause-enforceable-new-paper-out", "published_at": "2026-07-30 12:34:08+00:00", "updated_at": "2026-07-30 12:59:17.486519+00:00", "lang": "en", "topics": ["ai-policy", "ai-safety", "artificial-intelligence", "ai-research"], "entities": ["Raymond Koopmanschap", "Otto Barten", "How to Catch a GPU: A Taxonomy of Verification and Enforcement Mechanisms for International AI Agreements"], "alternates": {"html": "https://wpnews.pro/news/is-a-pause-enforceable-new-paper-out", "markdown": "https://wpnews.pro/news/is-a-pause-enforceable-new-paper-out.md", "text": "https://wpnews.pro/news/is-a-pause-enforceable-new-paper-out.txt", "jsonld": "https://wpnews.pro/news/is-a-pause-enforceable-new-paper-out.jsonld"}}