Milken Institute Global Conference

Michael Kratsios on why adoption wins the AI race

Michael Kratsios· Director of the White House Office of Science and Technology Policy at White House OSTP
·~18 min·English·Milken Institute
PolicyAI SafetyAI InfrastructureBusiness Strategy
TL;DR

The White House's top science advisor lays out how America keeps its AI lead: coordinating a decentralized government, opening classified military AI to every major vendor, using frontier models to harden cyber, regulating by use case, and winning through adoption at home and abroad.

01How the Role Works

No Ministry, So the White House Coordinates

<strong>The US has no science ministry</strong> — so the White House OSTP's job is to make State, the national labs, DARPA, and the Pentagon move in one direction.

what's very unique about the United States versus a lot of countries in the world is there isn't a tech or science um agency or ministry if you will. All of that work is spread out across a number of departments.

— Michael Kratsios, Milken Institute Global Conference
Key Insight
The org chart is the strategy. With no ministry and no direct budget authority, OSTP's only real lever is persuasion plus the annual R&D priorities memo — which is why 'coordinate' is the verb he keeps reaching for. Influence here is soft power dressed as administration.
02Defense Access

The Best Tools for the War Fighter

For classified military AI, the north star is <strong>the best possible tools for war fighters</strong> — so the department moved from reliance on one vendor to access from other major AI vendors.

the north star is we want to provide the best possible tools to our war fighters

— Michael Kratsios, Milken Institute Global Conference
Key Insight
The stated principle — 'best tools' — is also his answer to the Anthropic blacklist story. By reframing a values dispute (which lab will serve the military, on what terms) as a procurement-resilience problem, he lets the government avoid either endorsing or condemning any single company.
03Cyber Strategy

The Strongest Cyber Fortress in the World

A model called too dangerous to release becomes an asset: <strong>use it to build the strongest cyber fortress in the world</strong> by finding and patching vulnerabilities first.

the strongest cyber fortress in the world by using this model to identify these vulnerabilities to patch them and ultimately make them resilient to adversaries

— Michael Kratsios, Milken Institute Global Conference
Key Insight
He inverts Anthropic's framing. The very capability the lab called too powerful to release publicly is repriced, in his telling, as a one-time chance to out-harden adversaries. The model's danger doesn't go away — it gets reassigned as government advantage.
04Frontier Risk

Cyber Today, Bio and Nuclear Next

Cyber is <strong>the risk demonstrated today</strong>, while government needs capacity to evaluate biological, nuclear, and other domains as model capabilities improve.

Today's model, for example, may be really good at cyber and we want to be in a position to make that evaluation and be able to respond accordingly. But future models may be really good at things related to biological threats or maybe really good at nuclear related threats.

— Michael Kratsios, Milken Institute Global Conference
Key Insight
Kratsios is arguing for anticipatory, domain-specific evaluation capacity: cyber is the current proof point, but biological and nuclear testing must be built before future models expose those risks. The implication is that general AI oversight is insufficient without deep technical expertise across multiple domains.
05Regulatory Philosophy

Regulate the Use Case, Not the Technology

For AI applications, <strong>regulate by use case and sector</strong> — the FDA governs medical AI and the FAA governs drones.

if you are creating an AI powered medical diagnostic the FDA should be the one regulating that. If you are creating a commercial drone then the FAA should be regulating that.

— Michael Kratsios, Milken Institute Global Conference
Key Insight
Routing AI oversight through existing agencies is a deregulatory move wearing regulatory clothing. It avoids creating any new AI-specific authority or central gatekeeper, betting that regulators who already know their sector can absorb AI on top of what they do — a bet that only holds where those agencies are actually resourced to.
06Federalism

One Framework, Not Fifty

One national framework beats a 50-state patchwork because <strong>the patchwork quietly favors incumbents</strong> who can afford 50 sets of lawyers.

can we have one national framework where ultimately there's only one law and you don't have sort of individual states kind of running in in all these different directions and you provide some clarity and some certainty to innovators

— Michael Kratsios, Milken Institute Global Conference
Key Insight
The framing flips the usual politics of federal preemption. He sells one national law not as a favor to Big Tech but as protection for the little guy — the startup that can't staff 50 compliance teams. Left unstated is whether a single federal framework would actually be lighter-touch, or just easier for the largest players to shape.
07The Thesis

Adoption, Adoption, Adoption

Winning the AI race means <strong>adoption, adoption, adoption</strong> across government, industry, individual consumers, and the world.

the answer that I always give is adoption adoption adoption.

— Michael Kratsios, Milken Institute Global Conference
Key Insight
'Adoption' quietly redefines what winning means. Not the best model, not the most compute — market share of usage. That reframes the contest around diffusion rather than raw capability, a metric that conveniently favors whoever can ship to the most users, and get the rest of the world onto the American stack, first.
08Jobs & Entrepreneurship

AI Is the Entrepreneur's Unlock

He argues that <strong>AI is the biggest unlock ever for the American entrepreneur</strong>, making it easier for small teams to start disruptive companies.

There has never been a bigger unlock for the American entrepreneur than artificial intelligence. It has never been easier to start a disruptive company today.

— Michael Kratsios, Milken Institute Global Conference
Key Insight
Asked directly whether AI will drive layoffs into the unemployment numbers, he declines to forecast — 'I'm not a labor economist' — and pivots to entrepreneurship. The implied policy bet is that new company formation, not shielding existing jobs, is the administration's answer to AI-driven displacement.