Flash News

US Companies Jointly Oppose Restrictions on Open Source AI, Signatories Have Increased from 25 to 50

The U.S. is discussing whether to restrict downloadable and self-deployable open source models. Companies like Microsoft and NVIDIA subsequently co-signed an open letter urging against premature limitations.

The number of signatories has increased from 25 to 50, with OpenAI, Google, AMD, Cloudflare, GitHub, and Cisco joining later.

Buyers are vendors hoping to continue advancing the open model ecosystem and local deployment needs, while sellers are regulators who may tighten download and deployment permissions; the focus of this struggle is not the closed-source flagship itself, but the foundational card of open weights.

OpenAI and Google both retain their closed-source paid mainlines while each has open model products. Anthropic did not sign, consistent with its emphasis on distillation, leakage, and security risks.

Source: Public Information

ABAB AI Insight

The historical context of this game is clear: the U.S. is not discussing restrictions on open source AI for the first time; there were legislative attempts to strengthen export controls on open source models as early as 2024.

In terms of capital pathways, companies like OpenAI and Google are effectively implementing a "dual-track system": closed-source models continue to sell high-margin APIs and enterprise services, while open models are used to maintain the ecosystem, attract developers, and resist regulation. This is not a contradiction, but a layered defense.

In contrast, Anthropic seems to follow a "safety-first, controlled distribution" approach, prioritizing risk control over market expansion; while NVIDIA and Microsoft are more concerned about whether open models can continue to drive demand for computing power, software, and enterprise local deployments.

Structurally, this represents a change in regulation, but fundamentally it is a struggle for platform power: whoever can define whether a model is downloadable, deployable, and redistributable can determine the entry and boundaries of the AI ecosystem.

ABAB News · Cognitive Laws

  1. Open source is the entry point, closed source is the profit.
  2. Regulation takes the initiative, the ecosystem is rewritten first.
  3. True control lies in the distribution rules.

Source

·ABAB News
·
2 min read
·1d ago
分享: