Michael Kratsios on AI Policy: Flexible Rules and Startup Support

Michael Kratsios, Director of the White House's Office of Science and Technology Policy (OSTP) and the President's top advisor on science and technology, recently spoke at a Y Combinator startup school event. Kratsios, who previously served as the country's Chief Technology Officer during President Trump's first term and as Chief Operating Officer of Scale AI, discussed the intersection of AI policy and government, offering insights from both the private sector and the White House.

The Motivation for Public Service

Kratsios emphasized his fundamental belief in American leadership in emerging technologies, particularly AI. He views ensuring a regulatory environment that fosters success in AI as crucial for the American people to benefit from these advancements. He also noted that the U.S. government offers an unparalleled opportunity to work on problems of immense scale, a challenge he finds deeply rewarding. His goal is to guide the government to support, rather than hinder, technological progress.

Early Fascination with Technology and the Path to Policy

Kratsios's interest in technology began early, marked by his avid following of Steve Jobs' keynotes and the excitement surrounding the first iPhone's release. Although not an engineering student, he found his career path working with Peter Thiel for nearly seven years. During this time, he observed that regulatory issues consistently arose for companies across various industries, from ride-sharing services like Lyft to aerospace ventures like SpaceX. This experience highlighted how government policy could either unlock or impede technological innovation. When he joined the Trump administration, his initial focus was on streamlining regulations to facilitate innovation, such as enabling commercial drone operations and autonomous vehicles.

The Federated Nature of Technology Policy

A surprising aspect of technology policy-making in Washington, according to Kratsios, is its highly federated nature. Unlike health or defense, there isn't a single "Department of Technology." Instead, tech issues are distributed across multiple agencies, encompassing national security, commercial interests, and scientific research. The White House's role is to convene these diverse agencies to forge coherent policy. For instance, developing rules for commercial drone operations requires coordination between the Federal Aviation Administration (FAA) and agencies responsible for national security, ensuring drones don't pose risks to sensitive sites.

The White House Stance on Open Source AI

Addressing recent discussions and concerns about potential restrictions on open-weight AI models, Kratsios clarified the White House's position. He stated that the U.S. strategy for artificial intelligence, released in July of the previous year, explicitly commits to an open-source ecosystem. This strategy, co-authored by Kratsios, emphasizes the need for both vibrant closed and open-source environments for American leadership in AI. He views public discourse and input from the startup community as vital for policymakers to make informed decisions.

Bridging the Perception Gap Between Washington and the Tech Community

Kratsios identified a significant gap between how Washington perceives AI and how the tech community views it. In Washington, AI is seen as a broad technology impacting everything from public perception and data center concerns to healthcare, job displacement, and economic growth. Policymakers must balance these diverse implications, including public apprehension about data centers, which contrasts with the tech community's understanding of the critical need for computational infrastructure. Input from the startup community is essential to ensure that policy decisions support innovation effectively.

Regulating Rapidly Evolving Technology

The challenge of regulating an industry that reinvents itself every six months is immense. Kratsios argued against setting rigid "red lines" in policy, citing the European Union's AI Act as an example. This act was finalized before the advent of large language models (LLMs) like ChatGPT, meaning current LLMs must comply with rules written for a different technological landscape. He also mentioned the Biden administration's attempt to set a hard computational threshold for AI disclosures, which proved difficult to sustain. The White House aims for flexible policies that can adapt to the rapidly changing frontier of AI, recognizing the difficulty of resetting established regulations.

Overrated and Underrated AI Risks

Kratsios highlighted two primary AI risks:

• Cyber Risk: The dual-use nature of AI models, where capabilities that could be used for nefarious purposes can also be invaluable for strengthening existing systems. This presents a trade-off in developing and deploying powerful AI.
• Biological Risk: While acknowledging that concerns about biological risks have been present for years, he believes they have been somewhat "overblown" to date. However, he stressed the importance of building robust infrastructure for testing and evaluating models as they advance.

Ensuring "Little Tech" Has a Voice

Kratsios acknowledged the dominance of large tech companies in policy discussions. He emphasized the importance of engaging with the startup community and highlighted mechanisms like requests for information (RFIs) that allow a wide range of stakeholders to provide input. He pointed to the administration's focus on preventing a "patchwork" of state-level AI regulations, advocating for a single national standard. This approach, he argued, benefits smaller companies that lack the resources of large corporations to navigate complex, varied regulatory landscapes. He also noted the bipartisan antitrust efforts against big tech, which began in the first Trump administration and continue today, as a way to champion "little tech."

Fostering Innovation and Reducing Regulatory Barriers

Kratsios asserted that the current administration is highly committed to reducing regulations. He described two categories of technologies:

• Born-Free Technologies: These are technologies, like the early internet, that emerge without existing regulations. The government's role is to be cautious about introducing new rules that could stifle innovation in these areas.
• Born-in-Captivity Technologies: These require government approval to be commercialized, such as commercial drone operations or supersonic flight. The government's focus here is to relentlessly remove regulatory barriers, like the FAA waiver needed for drone flights or noise limits for supersonic aircraft, to enable these technologies to reach the market.

A Day in the Life at the White House OSTP

Kratsios explained that the White House serves as the central hub for coordinating policy across various federal agencies. The OSTP is one of four policy councils (alongside national security, domestic policy, and economic policy) that manage policy processes. His team, comprising specialists in areas like quantum computing, civil nuclear energy, biotech, and space, brings together relevant agencies (e.g., NASA and the Department of Defense for space policy) to resolve issues. His daily work involves:

1. Stakeholder Meetings: Engaging with external groups, including the tech community, to gather input.
2. Policy Implementation: Driving specific policy initiatives, such as pushing agencies like the FAA to modify rules to facilitate commercial drone flight.
3. Public Engagement: Communicating the President's message and understanding the challenges faced by people across the country.

He also clarified the hierarchy of roles within the White House, noting that his title as "Assistant to the President" signifies a senior advisory position. Policy disagreements among agencies are escalated through different levels until they reach the President for a final decision, though the goal is to resolve issues at lower levels to conserve the President's time.

"A New Golden Age of Science" Report

Kratsios discussed the recently released 150-page report, "A New Golden Age of Science," inspired by a 1945 letter from FDR to his science advisor, Vannevar Bush. Bush's report, "Science, The Endless Frontier," established the framework for government funding of basic research for 70 years. However, the landscape has changed dramatically: in 1950, the federal government funded 70% of R&D, while today, the private sector and philanthropy fund 70%.

The new report, commissioned by President Trump, aims to re-energize the science ecosystem. A core thesis is that AI will transform scientific discovery across fields like material science, pharmaceuticals, and chemistry. The government, which spends $200 billion annually on R&D, must prioritize AI integration to ensure the U.S. produces world-leading research.

The report envisions a future with AI agents designing experiments, contracting robotic cloud labs, and settling results on a ledger. It also proposes concrete actions like fast grants, private leverage for prizes, and agency action plans. Kratsios emphasized that the government's role is to "shape the arena, not direct discovery." He encouraged founders to build the infrastructure for autonomous experimentation, including robotics and software ecosystems, which the government cannot or should not directly develop.

Beyond AI: The Future of Quantum Computing

Beyond AI, Kratsios believes quantum information science and technology, particularly quantum computing, will be a major focus for Washington in the next five years. He drew a parallel to the early days of AI in 2017-2018, when he worked to convince officials of its importance. President Trump's 2019 executive order prioritizing AI laid the groundwork for its subsequent explosion. Kratsios sees a similar moment for quantum computing, with fundamental scientific questions still needing answers. The President recently signed an executive order on quantum information science, setting an ambitious goal for the Department of Energy to build a scientifically relevant quantum computer by 2028.

Congressional Role in AI Policy

Kratsios noted that congressional involvement in AI policy is complex due to the rapid pace of technological change. Unlike executive orders, which can be revoked, laws are permanent, making early, rigid legislation potentially harmful. He urged Congress to pass laws that are pro-innovation, specifically advocating for:

• National Standard for AI: To prevent a confusing patchwork of state-level regulations that disproportionately burden startups.
• Intellectual Property (IP) and Name, Image, and Likeness (NIL) Protections: To address concerns from creators about how AI impacts their work. He suggested that while model outputs that infringe on existing IP (e.g., Mickey Mouse) should be clearly prohibited, the industry could also explore revenue-sharing models for creators who wish to license their IP for AI training. He believes the market for such licensing will mature over time.

Encouraging Public Service

Kratsios, one of the youngest individuals to hold his position, encouraged young technologists to consider public service. He acknowledged the bureaucratic challenges but emphasized the profound fulfillment and impact one can have on the country. He believes that creating an environment where startups can thrive and new technologies can be built is a deeply rewarding aspect of government work, serving fellow Americans. He extended an invitation for interested individuals to explore opportunities within the White House OSTP.