- Pritzker Fellows
- Current Fellows
- Sasha Baker
Sasha Baker
Head of National Security Policy at OpenAI
Biography
Sasha Baker is the Head of National Security Policy at OpenAI. Previously, Sasha served as the Acting Under Secretary of Defense for Policy, after being confirmed by the Senate as the Deputy Under Secretary in February 2022. Sasha joined the civil service as a Presidential Management Fellow at the Office of Management and Budget and held key roles across administrations over nearly two decades of government service, including as Special Assistant to the President and Senior Director for Strategic Planning at the National Security Council; National Security Advisor for Senator Elizabeth Warren of Massachusetts and Deputy Policy Director on her 2020 presidential campaign; and Deputy Chief of Staff to Secretary of Defense Ashton Carter. Sasha holds a Bachelor of Arts in Government from Dartmouth College and a Master in Public Policy from the Harvard Kennedy School. She is a member of the Council on Foreign Relations and a recipient of the Department of Defense Distinguished Public Service Medal with oak leaf cluster.
Seminars
"AI & National Security"
Artificial intelligence is rapidly becoming a factor in military power, intelligence, cybersecurity, diplomacy, and economic competition. Governments around the world are racing to understand what it means—and how to use it. Yet many of the most important questions remain unresolved.
This seminar examines how AI is changing national security, not only by creating new risks, but also by expanding the ability of governments to understand, deter, defend against, and respond to threats. Designed for students without a technical background, the seminar series will develop the tools needed to think rigorously about technology, strategy, institutions, and decision-making under uncertainty.
We will explore a series of core questions. What is AI, and what is distinctive about modern foundation models? Which national security applications are real, near-term, and operationally useful—and which are mostly hype? Which risks deserve the greatest attention, including misuse, accidents, civil liberties harms, strategic instability, concentration of power, and failures of human oversight? How will AI shape competition and cooperation among states, technology companies, and open-source communities? And what institutions, rules, norms, and operational practices can help societies capture the benefits of AI while reducing its most serious dangers?
The seminar series deliberately balances opportunity and risk. Early sessions establish technical foundations and strategic context. Middle sessions examine operational applications and failure modes. Final sessions focus on governance, crisis management, and the choices leaders may face during fast-moving AI-enabled incidents. By the end, students should be able to assess competing claims about AI and national security, distinguish plausible futures from speculation, and develop their own views about how democratic societies should navigate one of the most consequential technological transitions of the twenty-first century.
Seminars are off the record and open to current UChicago students only.
**NOTE: This seminar topic, originally scheduled for Monday, October 5 at 12pm, is now being held on Monday, October 12 at 12pm.
This opening session establishes a shared foundation for the rest of the seminar. What are modern AI systems, how do they work, why have capabilities advanced so quickly, and what should we make of forecasts about future progress? The session will introduce key concepts including machine learning, foundation models, generative AI, agents, and autonomy, with an emphasis on the level of technical understanding needed to engage seriously with policy and strategy questions.
We’ll also explore one of the seminar's central themes: uncertainty. Governments increasingly must make decisions today based on expectations about what AI may be capable of tomorrow. Understanding the competing forecasts about AI's future (and the assumptions underlying them) is therefore essential to understanding the national security challenges and opportunities that will be discussed throughout the seminar.
Key Questions
- What distinguishes modern foundation models from earlier forms of AI?
- Why have capabilities advanced so quickly?
- What are the key inputs to AI progress?
- How should policymakers think about uncertainty and forecasting?
Special Guest: Dean Ball (via Zoom), Head of Strategic Futures at OpenAI
**NOTE: This seminar topic, originally scheduled for Monday, October 12 at 12pm, is now being held on Monday, October 19 at 12pm.
Few questions in technology generate more disagreement than whether artificial general intelligence (AGI) is possible, how soon it might arrive, and what it would mean if it did. This session will examine the competing visions of AI's future, from those who see current systems as the beginning of a transformative technological revolution to those who believe progress will be slower and more limited. Rather than attempting to predict the future, the goal is to understand the assumptions behind different forecasts and explore their implications for military power, economic competitiveness, scientific discovery, and global politics. We’ll also consider which risks and opportunities deserve attention today, even amid deep uncertainty about where the technology is ultimately headed.
Key Questions
- What do people mean when they say AGI?
- Why do forecasts differ so dramatically?
- If AGI were achieved, what would it mean for military, economic, and political power?
- Which concerns are speculative, and which deserve attention today?
Having established a foundation in the technology and its possible trajectory, we turn to how governments are already using AI today. This session will examine AI as a tool for intelligence, military operations, cybersecurity, scientific research, logistics, and broader state capacity. Particular attention will be paid to distinguishing between genuinely transformative applications and those that are primarily aspirational or overhyped. The central question isn’t whether AI will matter for national security, but where it’s likely to create meaningful advantages and what organizational changes are required to realize those benefits.
Key Questions
- How is AI changing intelligence, military operations, cybersecurity, and state capacity?
- Which applications are genuinely transformative, and which are overhyped?
- What advantages accrue to early adopters?
- What organizational changes are required to realize those advantages?
*This seminar will take place on Tuesday, October 27 at 3:30pm.
Powerful technologies create new vulnerabilities as well as new capabilities. This session explores the ways AI systems can fail, be manipulated, or produce unintended consequences in high-stakes national security environments. We’ll examine issues including reliability, adversarial attacks, misuse, model proliferation, automation bias, and failures of human oversight. The objective is to understand how technical failures can become strategic problems and how governments can evaluate, monitor, and manage risks when deploying increasingly capable AI systems.
Key Questions
- What are the most important technical and operational failure modes?
- How should governments think about misuse, accidents, and adversarial manipulation?
- What risks arise from increasingly capable and widely available models?
- How do leaders make decisions when they cannot fully trust the systems they are using?
As AI systems become more capable and more widely deployed, governments face difficult questions about how to maximize benefits while managing risks. This session explores the institutions, policies, and governance mechanisms that may shape the development and use of advanced AI systems. We’ll examine the respective roles of governments, technology companies, international organizations, and civil society, as well as the tradeoffs between innovation, security, accountability, and individual liberty. The focus isn’t on finding perfect solutions, but on understanding the practical choices policymakers face in a rapidly changing technological environment.
Key Questions
- What should governments regulate, and what should they encourage?
- What responsibilities belong to states, companies, and international organizations?
- How should democratic societies balance innovation, security, liberty, and accountability?
- What governance approaches are most likely to endure as the technology evolves?
Special Guest: Jason Matheny, President & CEO of RAND
AI isn’t only a technological development; it’s also reshaping international politics. This session examines how AI affects competition and cooperation among states, with particular attention to the evolving relationship between the United States and China. Topics include semiconductors, compute, export controls, talent, energy, industrial policy, open-source models, and the role of private firms in national power. We’ll also consider how AI may affect allies, developing countries, and the broader distribution of power in the international system. The goal is to understand whether AI represents a new arms race, a new industrial revolution, or something altogether different.
Key Questions
- Is AI best understood as an arms race, an economic competition, or something else?
- How should we think about U.S.-China competition in AI?
- What role do semiconductors, compute, talent, energy, and data play in national power?
- How might AI affect allies, developing countries, and the global distribution of power?
- Can international cooperation keep pace with technological change?
Special Guest: Rush Doshi, former Deputy Senior Director for China & Taiwan on the National Security Council at the White House
*This seminar will take place on Wednesday, November 11 at 3pm.
The final session places students in the role of senior policymakers confronting a fast-moving, AI-enabled national security crisis. Working under conditions of uncertainty and time pressure, students will be asked to assess incomplete information, balance competing priorities, coordinate across institutions, and develop recommendations for national leaders. The exercise is designed to integrate the major themes of our earlier discussions (technology, strategy, risk, governance, and geopolitics) while highlighting the challenges of decision-making when the stakes are high and certainty is impossible.