On July 28, 2026, Anthropic CEO Dario Amodei finally stepped into the spotlight to respond to the controversy surrounding the safety of open weight models. According to a single source, he first denied a widely circulated claim from the outside — that Anthropic has never advocated for an outright ban on open models. Instead, he emphasized that open weight models allow developers to freely use and modify the models, significantly reducing usage costs and enhancing the utility of the tools. The real disagreement was purposely shifted back to the “regulation pathway”: according to a single source, Dario suggested implementing “precise restrictions” starting from chip supply and model distillation processes, framing regulatory measures around computational power and technology diffusion, rather than simply treating open models as a risk source that should be banned outright. Prior to this, companies like OpenAI, Google, and SpaceX had joined a public open-source coalition around open weight models, and Anthropic became the only major cutting-edge model company that did not sign. This public response not only clarified its position but also starkly contrasted it with the signatories on the “model ban” versus “chip-distillation regulation” choices.

Refusing to Sign the Open Source Coalition: Anthropic Chooses Another Path

In the heated debate about the safety of open weight models, this open-source coalition was placed at the center of the industry. It focused on issues like “how open weights should be regulated” and “what abuse risks arise from the decentralization of model capabilities,” attempting to address the concerns of regulators and the public regarding open models, as outlined in public materials. Major tech companies such as OpenAI, Google, and SpaceX subsequently signed, and in the public opinion arena, it was seen as a “mainstream route”: advocating for a redefined line between safety and openness, aiming to set a more cautious framework for future policies. Since existing materials did not disclose specific terms and policy details of the coalition, this article can only describe its demands at this abstract level, without making technical or legal inferences about potential institutional designs.

Against the backdrop of this “mainstream route,” Anthropic became the only major cutting-edge model company that did not sign, subsequently being labeled as an “outsider.” Some supporters interpreted this as being more friendly to the open ecosystem, while others questioned whether it was “going off track” on safety issues. Dario then publicly emphasized that the company has never advocated for an outright ban on open models and recognized the potential of open weight models to reduce costs and unleash positive effects. According to a single source, this made the act of refusing to sign seem more like a rejection of the binary framework posed by the coalition — unwilling to compress complex risk issues into the simple option of “should open source be banned outright,” but rather insisting on shifting the regulatory focus to more specific technical aspects such as chip supply and model distillation.

Opposing an Outright Ban: The Costs and Red Lines of Open Models

In this response, Dario first clarified his stance: open weight models are not the problem itself; an outright ban might be. He emphasized that open weights could significantly lower the costs of usage and deployment, allowing more teams to create products and conduct research without relying on the closed interfaces of a few large companies. According to a single source, this constitutes what he termed “clear positive effects.” Open weights mean developers can freely call, fine-tune, and even transform underlying models, which is regarded in the industry as a key condition for accelerating iteration, boosting toolchains, and enriching community ecosystems. It is also the only gateway for many small and medium enterprises and academic institutions to engage in cutting-edge technology.

However, Dario did not shy away from addressing the real controversies surrounding the misuse of open models. Regulatory and research circles have long been vigilant about the potential for open weights to be used to generate fraudulent content, automate attacks, or facilitate high-risk applications that are more difficult to hold accountable. This concern has become part of the public discussion. Dario publicly expressed his opposition to a “one-size-fits-all” ban on open models. According to a single source, he hopes that while acknowledging that openness brings innovative and inclusive value, the safety issues can be broken down into manageable parts, drawing red lines between “openness” and “safety” through other forms of controls. Therefore, Anthropic neither joins the simple ban camp nor views openness as a benevolent technology that requires no constraints. Instead, it positions itself in the narrow space between the two: needing to leave developers enough room while also recognizing that this space must have boundaries and conditions.

Installing Controls at the Chip and Distillation Level: A Shift in Regulatory Pathways

In this “intermediate space” position, Anthropic's specific proposal is not to ban open weight models per se, but rather, according to a single source, to shift regulatory gates upstream and midstream: one end limiting the supply of critical chips, and the other end limiting the paths for model distillation. Distillation is a very common technology in the industry — compressing the capabilities of large models into smaller models, allowing the latter to inherit the “smartness” of the former while lowering costs and making deployment easier. If open weights can lower barriers, then distillation serves as an accelerator, bringing this low barrier into full bloom in edge devices and lightweight services. Dario's approach is to avoid a blunt “ban on open source” and instead focus on more specific issues like “who can access what computational power and how much frontier capability can be compressed out,” making abstract security debates fall on more executable regulatory points.

This set of claims did not emerge in a vacuum. Traditional finance and research institutions have regarded chip supply as a highly contested battleground: Andy Wong, head of multi-asset at Pictet, pointed out, according to a single source, that one of the focal points of current market debates is whether storage chips are taking too much profit, mentioning that the market hopes to see certain factors arise to reverse the perception that “SK Hynix is extracting too much profit from the supply chain”; meanwhile, Samsung Securities analyst Lee Jong-wook believes the progress of China's DUV has limited impact on this round of the AI chip cycle, and that the sell-off of semiconductor stocks in the US is an overreaction. According to a single source, these disagreements show that the narrative surrounding “who controls computational power and who profits from it” has long been highly politicized and cyclical. Anthropic's attempt to shift the regulatory focus to chip supply and distillation capabilities actually embeds the controversy of open models into this existing narrative of computational power competition, redrawing the boundaries between openness and risk at an already conflict-ridden entry point. In this context, installing controls at the chip and distillation level essentially redefines the rights of using computational power and capability, rather than simply declaring the fate of open models.

Divisions in the Open Source Camp: From Model Bans to Computational Power Regulation

The debate around open weight models is pushing the industry towards two distinctly different routes: one camp leaning towards “strong regulation” and the other emphasizing openness while advocating for “precise restrictions.” The former is represented by companies such as OpenAI, Google, and SpaceX that have already participated in the coalition. According to a single source, they have signed a coalition document regarding open weight models, seen in the public opinion arena as hoping to set clearer, potentially sterner boundaries for open models. In contrast, Anthropic did not sign the coalition among major cutting-edge model companies, but Dario Amodei publicly emphasized on July 28 that the company has never advocated for an outright ban on open models, recognizing the positive effects of open weight models in reducing costs and promoting the practical application of research and development, while also suggesting, according to a single source, that restrictions should focus on chip supply and model distillation, shifting regulatory discussions from “whether to allow open weights” to “how computational power and capabilities are disseminated.”

This differentiation is rewriting the future challenges of the open-source ecosystem. In the past, the open-source community sought to have model weights as open as possible so that more developers could engage in secondary development and experimentation within limited budgets. Now, if regulatory focus shifts to chips and distillation as per Anthropic's proposal, open-source projects will face not just the issue of accessing weights but also whether they can secure sufficient computational power within compliance boundaries, and whether a whole set of compliance costs and technical thresholds will need to be added around the distillation and retraining processes. There is a consensus that open weight models are closely related to the contributions of the open-source community, but under the narratives of strong regulation versus precise limitation, whether open source can maintain the same low-barrier widespread participation as before will likely depend on whether regulation ultimately chooses to constrain the models themselves or the chips and distillation pathways that carry model capabilities.

Chip Blockades and the Future of Open Source: The Three Major Uncertainties Ahead

As of Dario Amodei's public statement on July 28, 2026, the differences between Anthropic and other cutting-edge model companies regarding the open model debate have become manifest: on one side, regulators tend to respond to the uncertainties brought by open weights using a “one-size-fits-all” ban framework, while on the other, Anthropic emphasizes the costs and innovative dividends of open models, advocating to tighten controls on chip supply and model distillation processes, replacing outright bans with “precise limitations on computational power and dissemination.” However, existing public materials have not provided any technical details or timelines for chip or distillation restrictions, making the next three major uncertainties particularly critical: first, whether regulators will adopt this “precise limitation” approach instead of continuing with the vague model ban framework; second, how the chip industry will respond to potential computational power controls in the context of highly politicized debates on storage chip profit distribution, DUV progress, and cycle evaluation; third, how the naturally decentralized computational power and community cooperation-oriented open-source camp will reconstruct participation methods within potentially tightening compliance boundaries. For encryption and decentralized technology, the mid-term significance of this game lies not in the price of individual projects but in how the interfaces between computational power, models, and compliance will be delineated in the coming years — until these three uncertainties are resolved, decentralized computational power and AI-related on-chain projects will remain in a policy gray area that has yet to be distinctly defined.

Join our community, let's discuss and grow stronger together!
AiCoin exclusive Hyperliquid benefits: https://app.hyperliquid.xyz/join/AICOIN88
AiCoin exclusive Aster benefits: https://www.asterdex.com/zh-CN/referral/9C50e2
On-chain Telegram community: https://t.me/AiCoinWhaleData
On-chain community: https://www.aicoin.com/link/chat?cid=N6OVMor5g
AiCoin on-chain Twitter: https://x.com/aicoinwhaledata

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。