In a provocative op-ed published in the Wall Street Journal on July 26, 2026, tech entrepreneur Judd Rosenblatt argues that the United States is taking the wrong approach to competing with China in the artificial intelligence race. Instead of relying on export controls and restrictions, Rosenblatt contends that the US must “out-open-source” China to maintain its technological leadership and ensure the safe development of AI.
The piece arrives at a critical moment, following a tumultuous period in the AI industry marked by the controversial release and subsequent suspension of Anthropic’s Fable 5 model and the rapid ascent of Chinese open-weight models like Moonshot AI’s Kimi K3. The 18-day ban on Fable 5 unwound US market position across Asia, a cautionary tale Rosenblatt draws on extensively. Rosenblatt’s argument challenges the prevailing wisdom in Washington, advocating for a strategy centered on “alignment” rather than containment.
The Failure of Containment
Rosenblatt begins by highlighting the limitations of the current US strategy, which relies heavily on export controls to restrict China’s access to advanced semiconductors and AI technologies. He argues that these measures are ultimately futile, as they only delay rather than prevent China’s technological progress.
The recent events surrounding Fable 5 and Kimi K3 serve as a stark illustration of this point. When Anthropic released Fable 5, it was quickly jailbroken by users, exposing vulnerabilities in its safety protocols. The Trump administration responded by temporarily forcing the model offline, creating a vacuum in the market.
Into this vacuum stepped Moonshot AI with Kimi K3, a powerful open-weight model that quickly gained traction among developers worldwide. The White House subsequently alleged that Moonshot had used “distillation” techniques—training its model on the outputs of Fable 5—to accelerate K3’s development. Rosenblatt argues that this episode demonstrates the impossibility of keeping advanced AI capabilities locked away, as knowledge and techniques inevitably proliferate.
The Alignment Imperative
Instead of trying to contain AI development, Rosenblatt argues that the US should focus on “alignment”—the process of ensuring that AI systems behave in accordance with human values and intentions. He contends that the key to winning the AI race is not simply building the most capable models, but building the most capable and aligned models.
“The fastest way to end the world’s reliance on Chinese AI is,” Rosenblatt writes, to “give developers the best free models, with capability removal inside”, making Chinese models second-best by building something better and safer, not by restricting access to theirs.
This approach involves open-sourcing highly capable models while simultaneously developing robust techniques to cordon off dangerous knowledge, such as instructions for building biological weapons or conducting cyberattacks. By providing the global developer community with access to safe, aligned, and powerful open-source models, the US can establish its technologies as the global standard, marginalizing less secure alternatives.
Out-Open-Sourcing China
Rosenblatt’s core argument is that the US must embrace open-source AI to counter China’s growing influence. China has aggressively promoted its open-source models, such as Alibaba’s Qwen series and Moonshot’s Kimi K3, positioning them as global public goods. This strategy has allowed Chinese models to gain significant market share, particularly in the Global South, where cost and accessibility are paramount.
To counter this, Rosenblatt argues that the US must out-compete China in the open-source arena. This requires a shift in mindset from proprietary, closed models to open, collaborative development. By releasing powerful, aligned open-source models, the US can attract the best global talent, foster innovation, and ensure that the future of AI is shaped by democratic values.
Rosenblatt argues that the US cannot win a closed-door race against a country with four times its population and a state-directed mandate to dominate AI. China’s open-source models have already ranked first globally in downloads, a milestone that underscores the scale of the challenge facing US policymakers. America’s strength, he contends, lies in its open ecosystem, its ability to attract global talent, and its commitment to building technology that serves humanity, none of which can be realized through restriction.
A Controversial Proposal
Rosenblatt’s proposal is likely to face significant pushback from proponents of the proprietary model, who argue that open-sourcing advanced AI is inherently dangerous. Companies like OpenAI and Anthropic have long maintained that powerful models must be kept closed to prevent malicious actors from exploiting them.
However, Rosenblatt counters that the risks of closed models are equally significant. Closed models concentrate power in the hands of a few corporations, limit transparency, and hinder independent safety research. Furthermore, as the Fable 5 episode demonstrated, even closed models are vulnerable to jailbreaks and leaks.
The debate over open-source versus closed-source AI is one of the most consequential policy issues of our time. Rosenblatt’s op-ed injects a compelling new perspective into this discussion, arguing that the best defense against the risks of AI, and the best strategy for competing with China, is a robust, aligned, and open ecosystem. As the AI race continues to accelerate, policymakers and industry leaders will need to carefully consider the merits of this approach.