{"schema_version":"newruntime-agent-readable-v0.2","type":"post","stable_id":"post:sequoia-open-model-dependency-paradox","slug":"sequoia-open-model-dependency-paradox","title":"Open Weights Can Still Create A Strategic Dependency","description":"Sequoia argues that Western AI builders increasingly depend on Chinese open models as deployment substrates, post-training teachers, and sources of synthetic data.","retrieval_nugget":"Sequoia argues that Western AI builders increasingly depend on Chinese open models as deployment substrates, post-training teachers, and sources of synthetic data. Open weights give a team control over deployment of a particular model version. Sequoia's argument is that this does not eliminate dependency on the next release, the teacher used for post-training, or the synthetic data that carries capabilities.","status":"published","published_at":"2026-08-03","updated_at":"2026-08-03","record_date":"2026-08-03","date_kind":"published_at","topics":["open-models","ai-adoption","post-training","risk"],"source_urls":["https://sequoiacap.com/article/americas-open-model-paradox/"],"visuals":[{"id":"sequoia-open-model-dependency-paradox","kind":"editorial-diagram","role":"hero","src":"https://newruntime.com/images/posts/sequoia-open-model-dependency-paradox.webp","alt":"Hand-drawn dependency map where frontier capabilities flow indirectly through foreign open models into Western applications, synthetic data, post-training, and agent products, while a direct licensed teacher path is missing.","caption":"Open deployment control does not remove dependency on the next model, the training lineage, or the legal path for capability transfer.","credit":"New Runtime synthesis from Sequoia Capital","source_url":"https://sequoiacap.com/article/americas-open-model-paradox/","generated_with":"gemini-3.1-flash-image","width":1600,"height":900,"legend":[{"label":"Indirect path","description":"Frontier capabilities can reappear through foreign open models used as bases, teachers, or synthetic-data sources."},{"label":"Builder layer","description":"Applications, post-training systems, evaluations, and agents become dependent on each new open release."},{"label":"Missing path","description":"Sequoia proposes controlled domestic training rights as a legal capability-transfer route."}]}],"routes":{"html":"https://newruntime.com/posts/sequoia-open-model-dependency-paradox/","markdown":"https://newruntime.com/posts/sequoia-open-model-dependency-paradox.md","json":"https://newruntime.com/posts/sequoia-open-model-dependency-paradox.json"},"source_format":"markdown","next_reads":[{"type":"topic","path":"/topics/ai-adoption/","reason":"Explore the ai adoption topic hub.","url":"https://newruntime.com/topics/ai-adoption/","title":"AI Adoption - New Runtime","media_type":"text/html"},{"type":"topic","path":"/topics/open-models/","reason":"Explore the open models topic hub.","url":"https://newruntime.com/topics/open-models/","title":"Open Models - New Runtime","media_type":"text/html"},{"type":"related_material","path":"/posts/fireworks-lora-fullft-three-test-protocol/","reason":"Shares open models and post training.","url":"https://newruntime.com/posts/fireworks-lora-fullft-three-test-protocol/","title":"Run Three Tests Before Replacing LoRA With Full Fine-Tuning","media_type":"text/html"},{"type":"related_material","path":"/posts/arcee-open-model-science-post-training/","reason":"Shares open models and post training.","url":"https://newruntime.com/posts/arcee-open-model-science-post-training/","title":"Arcee Turns Scientific Post-Training Into A Run Ledger","media_type":"text/html"},{"type":"related_material","path":"/posts/open-weight-cyber-gap/","reason":"Shares open models and risk.","url":"https://newruntime.com/posts/open-weight-cyber-gap/","title":"Open-Weight Models Caught Frontier Cyber Capability with a Four-Month Lag","media_type":"text/html"}]}