Lennart Finke @finke.dev · 04/09/2026New blog post, about growing the European economy → fi-le.net/europefi-le.netAbundance in Europe - fi-le.netEurope is a great idea, an idea that I love. Let us think through how to contribute to its prosperity. 000
Lennart Finke @finke.dev · 21/07/2026Read it here → arxiv.org/abs/2607.16224arxiv.orgInternational Agreements to Limit Frontier AI: Objectives and ExitAn international agreement to limit AI development could be crucial to mitigate risks from AI. However, it remains unclear which conditions should determine when the limiting measures are relaxed.... 000
Lennart Finke @finke.dev · 21/07/2026Thanks to @aaronscher, @DavidSKrueger, @MauricBaker, @OliverEGuest, @matthijsMmaas, @thlarsen, @So8res for comments! Mistakes and bad takes are mine. 100
Lennart Finke @finke.dev · 21/07/2026This was a poster at @taig_icml, the Technical AI Governance workshop at @icmlconf. 100
Lennart Finke @finke.dev · 21/07/2026What happens when a state doesn't want to be in it any more? The Vienna Convention on the Law of Treaties permits no possibility of withdrawal (like in the UN Charter). But a withdrawal condition with 4 months notice (fastest datacenter buildout) could increase buy-in. 100
Lennart Finke @finke.dev · 21/07/2026The Vienna Convention on the Protection of the Ozone Layer offers inspiration: We could first designate the restrictions on AI development, and leave the relaxation conditions to be specified in a fixed time. A new IAEA would eventually verify them. 100
Lennart Finke @finke.dev · 21/07/2026It would be ideal to relax the treaty conditions based on progress in preparedness (alignment, resilience of society et cetera) but that is messy to determine. A committee that decides when the conditions are met would be scarily powerful. 110
Lennart Finke @finke.dev · 21/07/2026There's no historical precident for this: International security treaties (say the Biological Weapons Convention) are generally written to last forever. They sometimes offer a voluntary exit (like the Washington Naval Treaty), so that the treaty can "fail gracefully". 100
Lennart Finke @finke.dev · 21/07/2026New short paper: If we make an international treaty to limit AI development, when should it end? We want the benefits of AI eventually, but need time to prepare the world for it. It's an unsolved problem how the treaty ends at the right time by design. 120
Reposted by Lennart FinkeCas (Stephen Casper) @scasper.bsky.social · 20/07/2026🚨 New paper: Some, but not all, AI companies make corporately-loyal models. xAI, DeepSeek, Anthropic, & OpenAI models all downplay company controversies. Google, Meta, & Alibaba models don't. The findings are clear, but we are pretty confused as to why... 🧵 @finke.dev 2186
