Michael Kratsios is the Director of the White House Office of Science and Technology Policy (OSTP) in the Executive Office of the President (2025–present). He served as US Chief Technology Officer and OSTP Deputy CTO in the first Trump administration (2017–2021) and as Managing Director at Scale AI between administrations (2021–2025), making him a continuity figure across the two Trump AI agendas. As OSTP Director he is the senior science-and-technology adviser to the President and the principal executive-branch figure below the cabinet level on AI research, workforce, and standards policy.
Background
In the first Trump administration, Kratsios served as OSTP Deputy CTO and later as US CTO, and was a principal author of Executive Order 13859 (the 2019 "American AI Initiative"). Between administrations, he served as managing director at Scale AI, a major supplier of evaluation and data infrastructure to US frontier labs and to the federal government, including the CDAO and DIU. In January 2025 he was nominated and subsequently confirmed as OSTP Director.
Roles and AI policy work
First term: EO 13859 and the American AI Initiative
As OSTP Deputy CTO and later US CTO, Kratsios was a principal architect of EO 13859 in February 2019 and the subsequent American AI Initiative strategy, which set out early federal arrangements for AI R&D coordination, workforce, and international engagement.
Second term: EO 14365 and the AI Action Plan
Kratsios is one of the senior officials responsible for implementing EO 14365 and America's AI Action Plan across federal science and technology agencies. OSTP serves as a coordinator for federal R&D allocation, NIST standards direction, and interagency AI workforce efforts.
NSTM-4 on adversarial distillation
On April 23, 2026, Kratsios signed NSTM-4, "Adversarial Distillation of American AI Models," as OSTP Director (OSTP NSTM-4: Adversarial Distillation of American AI Models). The memo alleges that foreign entities, primarily China-based, are running coordinated industrial-scale campaigns using tens of thousands of proxy accounts and jailbreaking techniques to distill American frontier models, and commits the administration to share threat intelligence with US AI companies, develop counter-distillation defenses, and explore foreign-actor accountability. The broader concept and framework are covered at Adversarial Distillation and Adversarial Distillation. The memo was issued the same week as a State Department cable on alleged DeepSeek IP theft, against the backdrop of the US-China posture surrounding the DeepSeek V4 launch.
July 2026: Moonshot allegations and the "Golden Age" science report
On July 22, 2026, Kratsios said the administration has "information that Moonshot AI distilled Anthropic's Fable for the development of its K3 model," and separately said Moonshot acquired GB300-equipped servers and accessed the banned Nvidia chips in Thailand, likely for training (Source: x.com; bloomberg.com). In remarks made public July 23, he sought to distinguish Chinese "distillation operations" — models he says benefit secretively from American frontier developers' work — from open-source AI generally, as Chinese open-weight models gained popularity among U.S. companies (Source: insideaipolicy.com). The Hill reported that Kratsios accused Moonshot of building an internal platform for large-scale distillation against Anthropic's Fable model, as the administration reportedly revived a push to ban Chinese AI models over objections from figures including Sriram Krishnan and David Sacks (Source: thehill.com; fastcompany.com). See Distillation, Chip Smuggling and Export-Control Evasion.
On July 21, 2026, Kratsios released "Science: A New Golden Age", presented as the first comprehensive rethinking of the federal science enterprise since Vannevar Bush's 1945 report, shifting support toward individual scientists and ARPA-style programs, making AI-accelerated research a core priority, and requiring agencies with research budgets of $3 billion or more to submit action plans within 90 days (Source: whitehouse.gov; insideaipolicy.com).
Scale AI tenure (2021–2025)
Kratsios' tenure at Scale AI placed him at a supplier of evaluation and data infrastructure used by both US frontier labs and the federal government, including the CDAO and DIU, giving him direct exposure to the operational mechanics of frontier-model evaluation and government AI contracting.
Kratsios was among those under consideration for the AI czar post Trump announced on September 19, 2026, alongside Treasury Secretary Scott Bessent, reported on September 22 as the frontrunner, and OPM Director Scott Kupor; the decision was not final (Source: semafor.com).
G20 innovation ministerial (September 2026)
Kratsios co-hosted a two-day G20 innovation ministerial in Chapel Hill, North Carolina, and on September 1, 2026 urged delegates to adopt what he called the "Carolina Principles," under which signatories would agree to avoid writing entirely new AI regulations and instead write rules only for "novel" situations involving the technology. He told reporters that the Chinese government, represented by Technology and Science Minister Yin Hejun, had signed, but did not provide a copy of the document. Other speakers took differing positions: Google DeepMind's Demis Hassabis, addressing the meeting by video, called on the G20 to set up safety tests for AI systems; Meta's Mark Zuckerberg said countries should not restrict open-weight models; and SpaceX's Elon Musk said EU policy "inhibits progress," cited a likely 15-gigawatt power shortage for AI chips in the following year and urged countries to build data centers. Former White House AI adviser David Sacks derided the idea of an "FDA for AI" that would approve models before release. Anthropic co-founder Tom Brown, OpenAI's Sam Altman and Nvidia's Jensen Huang were scheduled to appear on September 2; the G20 leaders' summit is set for Miami in December 2026 (Source: reuters.com; insideaipolicy.com; theinformation.com). Appearing on September 2 alongside Commerce Secretary Howard Lutnick, Huang urged G20 countries to avoid AI regulations focused on "theoretical harms" and to write rules for real-world problems instead (Source: reuters.com).
Positions
Commentators describe Kratsios as more technocratic than David Sacks, generally pro-acceleration and skeptical of prescriptive regulation, with a posture oriented toward strengthening American R&D and standards infrastructure rather than aggressive deregulation, representing the more institutionally oriented wing of the administration on AI.
In the September 2026 pacing debate Kratsios rejected the idea that technology companies need government intervention to ensure the safety of their AI models, according to Business Insider, and in a Fox News interview on September 20, 2026 he said that AI laboratories that believe their products are dangerous could "just stop" (Source: businessinsider.com). See Coordinated slowdown proposals compared.
Kratsios carried the same position to the UN Security Council, in an appearance whose coverage became public on September 24, 2026. He sought to assure Council members that the United States is committed to responsible development and deployment of AI even as it shuns restrictive regulation, and used the appearance to promote the administration's goal of establishing the American AI stack abroad through its AI exports program, saying: "The frontier of intelligence is advancing rapidly. That is not a reason to pause its further development or to constrain it with new global governance structures." Only the lede of the coverage was retrieved (Source: insideaipolicy.com).
Sources
- Executive Order 13859 — Maintaining American Leadership in Artificial Intelligence and source summary
- Executive Order 14365 — Ensuring a National Policy Framework for AI and source summary
- America's AI Action Plan and source summary
Relationships
- related: Donald Trump — OSTP Director; senior S&T adviser.
- related: Carolina Principles for Emerging Technologies — G20 principles he promoted; SANDBOX Act (S. 2750) — bill that would make the OSTP director administrator of a federal AI sandbox
- related: Executive Order 13859 — Maintaining American Leadership in Artificial Intelligence — principal first-term author.
- related: Executive Order 14365 — Ensuring a National Policy Framework for AI, America's AI Action Plan — senior second-term implementer.
- related: David Sacks, JD Vance — peer second-term AI policy leadership.
- related: National Institute of Standards and Technology (NIST) — OSTP coordinates federal standards work, including NIST AI activities.