The White House named 12 AI subfields. Open weights was not one of them.

The Office of Science and Technology Policy has published a National Security Science and Technology strategy, dated August 2026, naming 15 critical technology areas the United States intends to lead, with artificial intelligence and autonomy among them and 12 subfields listed beneath it ^[1][2][3]. Open-weight AI is not one of the 12, and neither open weights nor open-source software development appears anywhere in the 24-page document ^[4][5].

The list is not decorative. The strategy meets a requirement written into the CHIPS and Science Act obliging OSTP to submit a science and innovation strategy after each National Security Strategy, this one supporting the 2025 NSS, and its appendix updates the official OSTP list of critical and emerging technologies ^[6][7][8].

Here is what did get named under AI and autonomy: perception and sensor fusion, planning and reasoning, robotics and embodied intelligence, foundation models, multi-agent systems and swarm intelligence, autonomous systems across five domains, autonomous command and control, agent identification and authentication, interpretability and control, adversarial robustness and AI security, continual learning, and distributed and privacy-preserving machine learning ^[9][10]. Foundation models are defined as "large language, multimodal, and world systems" ^[11]. Nothing in the entry refers to model weights, open release, or compute thresholds ^[12]. The word "open" appears five times in the whole strategy, each time describing open research environments or an open investment climate, never a model ^[13].

Two nearby entries make the gap easier to see. The strategy calls brain-computer interfaces critical under future computing technologies, and distributed ledger technologies and digital assets critical under information management and cybersecurity ^[14][15]. So blockchain is on the list and open weights are not ^[16].

The procurement passages point the other way on cost. The strategy says the United States will seek "optimal combinations of lower-cost (and sometimes lower-tech or attritable) platforms" that can be fielded in larger numbers ^[17], pushes Pentagon buyers towards other transaction authorities and milestone-based contracts ^[18], and tells agencies to try "faster non-traditional approaches that may involve higher-risk but offer high potential reward" ^[19]. Attritable hardware, in other words, alongside silence on the cheapest layer of the software stack.

Patrick Tucker at Nextgov first flagged the omission and called it a huge oversight, arguing that the strategy sides with a handful of well-connected companies and their energy-intensive approach to AI while cheaper alternatives that researchers increasingly favour lose out ^[20][21]. The industry itself had already split on open weights in July, with Nvidia on one side and OpenAI and Anthropic absent from the other ^[22]. Tucker cites Michael Schiffer, a senior advisor at Scalare Advisors and senior fellow at the Center for American Progress, who argued in Just Security on 18 August that export controls can no longer decide who leads in AI, writing that Chinese models took 41% of Hugging Face downloads over the past year and that Alibaba's Qwen family alone has spawned more than 200,000 derivative models ^[23][24][25]. One caveat on that chain of reasoning: Schiffer's essay never mentions the strategy and is a general argument about American technology policy, which Nextgov applied to the document ^[26][27].

The strategy also argues against reading the omission as a decision. Each critical technology area "includes a set of subfields that illustrate its scope but are not meant to be comprehensive", and the appendix is offered as an update to an existing list rather than a closed one ^[28][29].

What to watch is the next document, not this one. OSTP and the National Security Council have promised technology-specific strategies "at the next level of detail" for the technologies in the appendix, so an AI plan is still to come and open weights could appear in it ^[30][31]. Until then, the follow-on work is scoped by an appendix that does not name them.