America’s Open-Model Paradox

2026-07-24 · Source: Sequoia Capital · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Public Policy & Governance, Cybersecurity & Data Privacy · Depth: Intermediate, medium

Summary

Western AI companies are increasingly dependent on Chinese open-source models, such as Qwen and Kimi, for fine-tuning, synthetic data generation, and as foundational components. Qwen's share of new open-model fine-tunes surged from 1% in January 2024 to 69% by February 2026, indicating a significant shift. This reliance stems from legal restrictions on using outputs from American frontier models like GPT or Claude for similar purposes, compelling Western labs to utilize lawfully accessible Chinese alternatives. This dynamic grants Chinese labs a structural advantage, as they can distill capabilities from Western frontier advances, often through alleged unauthorized extraction, and then release open weights that Western builders can legally adopt. The situation presents supply-chain risks, as open weights do not guarantee the absence of embedded backdoors, a critical concern for defense and infrastructure. To counter this, a framework is proposed: continue developing Western base models, establish controlled teacher access from frontier labs to qualifying Western companies, and intensify efforts to raise the cost of foreign distillation.

Key takeaway

For Directors of AI/ML and VPs of Engineering evaluating foundational model strategies, your reliance on Chinese open-source models for cost-effective development presents significant long-term dependency and security risks. You should advocate for and explore structured access to American frontier model capabilities, such as controlled training rights, to build a resilient, auditable domestic AI supply chain. This shift is crucial to avoid falling behind if foreign access is restricted and to mitigate potential backdoor vulnerabilities in critical applications.

Key insights

Western AI development faces a paradox of relying on Chinese open models due to legal restrictions on domestic frontier model use.

Principles

• Legal frameworks dictate open model adoption.
• Open weights do not guarantee security.
• Frontier model access drives capability transfer.

Method

Establish controlled teacher access from frontier labs to qualifying Western companies, selling structured training rights for capability transfer.

In practice

• Evaluate supply-chain risks in open-weight models.
• Advocate for domestic capability transfer policies.
• Invest in Western base model development.

Topics

• Open-source AI models
• AI supply chain
• Model distillation
• Frontier AI
• National AI strategy
• AI security risks

Best for: CTO, Investor, Entrepreneur, Director of AI/ML, VP of Engineering/Data, Policy Maker

Related on AIssential

• Thoughts on "Open" Models — The debate over "open weight" AI models is driven by complex geopolitical, economic, and security interests among major tech players and governments.
• Arcee, a US open source AI lab, says Chinese models are not inherently dangerous — Arcee's CTO asserts Chinese open-weight AI models are not inherently dangerous, advocating for competition over bans.
• As Anthropic suspends access to new models, India debates its AI future — Geopolitical factors can abruptly restrict access to critical frontier AI models, highlighting technological dependence risks.
• The Once And Future Fable #5 — The "Mythos Moment" reveals an ad hoc, reactive US AI policy struggling with frontier model control, legal frameworks, and open-source risks.
• US to restrict American companies’ use of Chinese AI models — The US plans to restrict American companies' use of Chinese AI models due to national security, while China considers reciprocal restrictions.
• State of AI: April 2026 newsletter — The AI landscape is marked by rapid model advancement, geopolitical tension, and unprecedented commercial growth, challenging existing regulatory and ethical frameworks.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Sequoia Capital.