Scale AI Response to the Request for Information on the Development of an Artificial Intelligence Action Plan

What it argues for

This is Scale AI's eight-page submission to the White House Office of Science and Technology Policy's request for information on an AI Action Plan, signed by Max Fenkell, Head of Government Relations. It is an industrial-strategy document framed entirely around competition with China, not around AI risk. It opens by endorsing the RFI's framing, "Scale strongly supports President Trump’s mission", and builds its case on a diagnosis that the race is close: "From Scale’s perspective, today, China is leading on data, we are tied on algorithms, and the United States remains ahead on compute." From there it proposes action on "four main pillars – Protect, Promote, Unleash, and Innovate". Protect means reviewing existing export controls before adding new ones and enforcing those already in place, because "Export controls are only as effective if they are properly enforced". Promote means giving allies full access to US technology and extending the remits of the Export-Import Bank and the Development Finance Corporation to AI and software exports. It also means keeping NIST and a US-chaired Global Network of AI Safety Institutes at the centre of AI measurement science, with China kept out. Unleash means government adoption, starting at the Department of Defense with AI-ready data requirements and enterprise-wide data infrastructure, and an AI strategy focused on "Agentic Applications such as Agentic Warfare". Innovate carries its regulatory position: "Scale has long been vocal proponents of the idea that the right regulatory framework is one that maximizes innovation, but still creates proper guardrails". It would deliver that through "a sector-specific and use-cased base regulatory framework which rightly governs the use of the technology instead of the technology itself", plus a comprehensive regulatory gap analysis. It closes with a workforce argument that also describes Scale's own marketplace: "If our nation wants to build the highest quality AI systems in the United States, they must be trained by people in the United States."

Stated positions (13)

• Frames the whole question as a race with China: "The next four years will be crucial to deciding where the world’s premier AI systems will be built and whether they will be trained and shaped by Western or Chinese values."
• Diagnoses a near-tie on the three inputs to AI: "From Scale’s perspective, today, China is leading on data, we are tied on algorithms, and the United States remains ahead on compute." On data it cites Scale's own analysis that "China spent at least $1.2 billion dollars on data labeling alone".
• Protect, part one: no new export controls without a review first. "Scale recommends that prior to putting in place any new export controls, the Trump Administration does a comprehensive review of the USG’s existing controls to better understand where new ones are required."
• Protect, part two: enforce what exists. "Export controls are only as effective if they are properly enforced", and "the smuggling of powerful semiconductor chips into countries that should not have access to US technology is well known". It concedes that "export controls, alone, are not going to make certain that we beat China".
• Promote: allies asked to choose US over Chinese technology must get it. "the United States must also provide those countries full access to US technology to ensure effective alignment". It cites 5G as the cautionary case, where under-promotion led to "China effectively becoming the de facto global standard".
• Asks that "programs like the Export-Import Bank and US Development Finance Corporation" have their remits extended to include AI and software exports, stressing "this is not an increased funding recommendation but rather a recommendation to expand eligibility and scope to meet the realities of today."
• Wants US leadership in AI measurement science: "the US must also lead in the development of measurement science for AI including standards and frameworks", with NIST central. It calls the US-chaired Global Network of AI Safety Institutes "a ready-made entity to export our measurement science".
• Would keep China out of that network: "To date, China has not been allowed entry into the body and so long as the US has a seat at the head of the table it should be guaranteed to stay that way."
• Unleash: says the Department of Defense, though furthest along, "has not launched a new AI program in nearly a decade". It asks for "AI-ready data requirements and enterprise-wide AI data infrastructure", noting "Last year alone, the DOD spent under $100 million on AI-ready data whereas China spent over $1 billion."
• Proposes a defence AI strategy built on autonomy: "Scale believes that this should focus on Agentic Applications such as Agentic Warfare." It discloses its own stake: "Scale is actively working on deploying the first instance of this".
• Innovate: regulate use, not technology. It asks the Administration to reaffirm "a sector-specific and use-cased base regulatory framework which rightly governs the use of the technology instead of the technology itself", by analogy with a laptop: "we do not regulate the laptop, we regulate the malicious use of said laptop."
• Accepts that gaps may exist: "Scale strongly supports the Administration conducting a comprehensive regulatory gap analysis which would highlight if one exists. If shown to exist, actions must be taken to provide regulatory consistency."
• Workforce: "If our nation wants to build the highest quality AI systems in the United States, they must be trained by people in the United States." It points to its own marketplace, whose workers were "compensated nearly $500 million collectively" in 2024, and asks policy to "promote the AI upskilling of the American workforce by incentivizing business activities that enable workers to transfer their pre-existing credentials, skills and competencies to the AI economy".

About this document

An 8-page PDF rendered from Google Docs, on a letterhead reading 'Scale AI, Inc. | 650 Townsend St. | San Francisco, CA 94103' and headed 'SCALE AI RESPONSE TO THE REQUEST FOR INFORMATION ON THE DEVELOPMENT OF AN ARTIFICIAL INTELLIGENCE ACTION PLAN'. It is signed at the end by Max Fenkell, Head of Government Relations, Scale AI. It runs as continuous prose with 15 footnotes, mostly news links and Scale's own analysis. It opens with an introduction and a section on the race with China (data labelling spend, PLA contracts with LLM companies, DeepSeek). It then sets out four pillars, headed 'Pilar One: Protect', 'Pilar Two: Promote', 'Pilar Three: Unleash' and 'Pilar Four: Innovate' (the misspelling is in the original), and ends with a short conclusion. Its asks are addressed to the Administration: export-control review and enforcement, Ex-Im Bank and DFC eligibility for AI, NIST and AI Safety Institute network leadership, DOD data and agentic-warfare priorities, a use-based regulatory framework with a gap analysis, and workforce policy. Footnote 9 cites a Reuters piece on Microsoft's 2024 G42 deal as an example of a country choosing US technology.

How this sits against AI law

Each stance compared with what EU and US instruments actually require. Where no instrument addresses a theme, that gap is shown rather than hidden.

Much of what Scale asked for became federal policy four months later. America's AI Action Plan of July 2025 has a section on strengthening export-control enforcement, a full-stack AI export programme with the Export-Import Bank and the Development Finance Corporation named as coordinating agencies, an evaluations ecosystem under NIST's CAISI, a push for AI adoption inside the Department of Defense and a worker-skills agenda. The White House's March 2026 legislative recommendations then endorsed Scale's regulatory model: no new AI regulator, with sector-specific regulation through existing agencies. The two points where Scale and the plan differ are the gap analysis and the safety-institute network. Scale's gap analysis looks for missing regulation, while the plan's review looks for regulation to repeal. And the plan never mentions the Global Network of AI Safety Institutes that Scale wanted the US to lead. Against the EU the gap is one of design. The AI Act regulates AI systems by use case, which suits Scale's model, but it also places obligations directly on general-purpose models regardless of use, which is exactly the "technology itself" Scale says should not be governed.

Source

Date on the page:

3/15/2025

Source checked:

opened and confirmed on 2026-09-30