{"version":1,"type":"story","url":"https://digestai.news/story/rand-says-us-should-adopt-a-freedom-of-action-ai-strategy-to-keep-opti","json":"https://digestai.news/story/rand-says-us-should-adopt-a-freedom-of-action-ai-strategy-to-keep-opti.json","markdown":"https://digestai.news/story/rand-says-us-should-adopt-a-freedom-of-action-ai-strategy-to-keep-opti.md","slug":"rand-says-us-should-adopt-a-freedom-of-action-ai-strategy-to-keep-opti","headline":"Rand says US should adopt a ‘freedom of action’ AI strategy to keep options open","summary":"RAND’s new paper argues the United States should follow a “freedom of action” approach to AI, keeping all strategic options open while investing in a human‑AI ecosystem, an AI‑security architecture, modernising legacy security institutions, and boosting citizen, firm and government capacity. The authors list four ingredients for the strategy and outline seven archetypal approaches – from coexistence and denial to acceleration – together with five key uncertainties about danger proximity, coexistence feasibility, restraint feasibility, decisive advantage and suppression feasibility.\n\nA team of researchers reported creating xenocortical mice by transplanting human stem‑cell‑derived cortical organoids into neonates whose mouse neocortex and hippocampus were genetically depleted. The grafted human neurons integrated with mouse circuitry, showed organized activity, and gave the mice intermediate behavioural patterns in motion‑sequencing analyses. In a maze test the xenocortical mice performed above chance, suggesting the human tissue can support memory functions. The work is presented as a platform for testing medical therapies for human brains.\n\nSeparate analysis of uncensored open‑weight models found 3,471 HuggingFace repositories, with the top five model families being Qwen, Llama, Gemma, MistralMixtral and Phi. Each original model is repackaged on average 2.4 times; 10 actors control 45% of the repos. Chinese‑origin models now represent 38% of uncensored production, up from 1% in early 2024. The authors argue that tracking such redistribution helps anticipate how future autonomous systems might modify models.\n\nFinally, a new pacing‑research agenda, authored by scholars from ACS Research, University of Toronto, Harvard and others, outlines why deliberate pacing of AI progress may be needed, what could be paced (compute, data, model weights, deployment), and the trade‑offs of pre‑pacing, trigger, intervention and exit phases. Toby Ord’s paper on recursive self‑improvement is also summarised, describing four phases of an intelligence explosion and various physical and algorithmic limits that could curb singular growth.","keyPoints":["Rand recommends a “freedom of action” AI strategy for the US, stressing ecosystem, security, legacy overhaul, and citizen capacity.","Researchers created xenocortical mice with human organoid grafts that showed integrated neural activity and intermediate maze performance.","Analysis of uncensored models finds 3,471 HuggingFace repos, average 2.4 redistributions, and Chinese‑origin models now 38% of new production."],"whyItMatters":"The policy outline, new mouse‑human brain platform, and mapping of uncensored model ecosystems each shape future AI governance, medical research, and security risk assessments.","category":{"slug":"policy","name":"Policy & Regulation","url":"https://digestai.news/category/policy"},"entities":{"companies":["Rand","HuggingFace","10a Labs"],"models":["Qwen","Llama","Gemma","MistralMixtral","Phi"],"people":["Jack Clark","Toby Ord","Cédric Villani","Terence Tao"]},"firstPublishedAt":"2026-09-21T12:31:04Z","updatedAt":"2026-09-21T12:31:04Z","sourceCount":1,"hasPrimarySource":false,"sources":[{"outlet":"Import AI","title":"Import AI 473: The US’s superintelligence strategy; human brain in a mouse skull; and machine hermeneutics","url":"https://jack-clark.net/2026/09/21/import-ai-473-the-uss-superintelligence-strategy-human-brain-in-a-mouse-skull-and-machine-hermeneutics","publishedAt":"2026-09-21T12:31:04Z","type":"newsletter","primary":false,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Rand says US should adopt a ‘freedom of action’ AI strategy to keep options open\", 21 September 2026, https://digestai.news/story/rand-says-us-should-adopt-a-freedom-of-action-ai-strategy-to-keep-opti","publisher":"Digest AI","title":"Rand says US should adopt a ‘freedom of action’ AI strategy to keep options open","datePublished":"2026-09-21T12:31:04Z","url":"https://digestai.news/story/rand-says-us-should-adopt-a-freedom-of-action-ai-strategy-to-keep-opti"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}