Stanford Institute for Human-Centered Artificial Intelligence (HAI)

Public attitudes toward AI will shape how governments regulate it, how firms invest, and how quickly new tools spread through workplaces. Stanford researchers Michael Tomz and Diyi Yang are conducting the first systematic comparison of public opinion on AI in the United States and China — one of three projects receiving grants from Stanford HAI and the Hoover Institution's Technology Policy Accelerator. Learn about all three projects: https://lnkd.in/gfbBnRmm

Congratulations to Sanmi Koyejo, Chelsea Finn, and Azalia Mirhoseini on being named to the TIME AI 100, a recognition of innovators, leaders, and thinkers reshaping the world through advances in artificial intelligence. Working across some of the most consequential questions in AI, these three Stanford faculty are tackling human-centered AI problems spanning from how robots learn, how chips get designed, and whether AI systems actually do what we think they do once they are out in the world. Congratulations! ti.me/100ai

Every system built to detect child sexual abuse material online involves unavoidable trade-offs. Too sensitive, and innocent users are reported to authorities. Too lenient, and abuse goes undetected. Despite near-universal agreement that this material is illegal, there is no uniform framework dictating what providers must do. That means the hardest choices — how to configure detection systems, whether to use human reviewers, how to handle false positives — fall to providers and the computing professionals building these systems. Stanford HAI Policy Fellow Riana Pfefferkorn on the ethical weight of those decisions: https://lnkd.in/gG8wA2ik

What does America's AI future look like — and is Washington ready for it? That question was at the center of a remarkable conversation a few weeks ago at Stanford HAI, where NVIDIA CEO Jensen Huang, former 66th US Secretary of State and Director of the Hoover Institution Condoleezza Rice, and Stanford HAI founding director Fei-Fei Li joined congressional staffers for the keynote dinner of our Congressional Boot Camp on AI. A few moments that stuck with us: 🔹 "Anxiety is not a policy." — Condoleezza Rice, on the need for Washington to move from hesitation to concrete action 🔹 "Closed models is a business model choice. Open models are essential for human infrastructure." — Jensen Huang, making the case for open-source AI as foundational — not optional 🔹 "What's excessive in the policy circle right now is rhetoric fueled by fear and doomerism. We need a scientific, rational understanding of AI." — Fei-Fei Li The conversation covered everything from export controls and national security to open-source models. The agreement across the panel: the gap between political narratives and technological reality is real — and closing it matters enormously. That's exactly why Stanford HAI created the Congressional Boot Camp on AI. Over three days, staffers from both the House and Senate engage directly with leading scholars, technologists, and policy experts to build the knowledge needed to govern this technology thoughtfully.

Congratulations to HAI Founding Director Fei-Fei Li, who has been named to the Time AI 100! A true trailblazer in the world of AI, Fei-Fei is renowned for co-creating ImageNet, the groundbreaking visual database that helped revolutionize deep learning research, co-founding the Stanford Institute for Human-Centered AI (HAI), and tirelessly championing ethical, human-centered AI and smart policy. This recognition is a testament to her impact on the field and her unwavering commitment to ensuring AI benefits all of humanity. Congratulations, Fei-Fei!

Stanford HAI's AI Index is hiring a Research Scientist to help track, analyze, and communicate the state of AI. The role supports the annual AI Index Report — monitoring new research, collecting and validating data, consulting outside experts, and writing technical sections of the report. It also includes maintaining and enhancing the Global AI Vibrancy Tool. Learn more and apply: https://lnkd.in/g8qfqyp2

California's Delete Act requires data brokers to let consumers request deletion of their data and report how many requests they receive. A new Stanford report finds most aren't doing either. Stanford HAI Privacy and Data Policy Fellow Dr. Jennifer King spoke to Marketplace Tech:  https://lnkd.in/gNBswZfy 

AI agents can book your travel, manage your finances, and reschedule your doctor's appointments. The more data an AI agent can access, the more useful it becomes — but also the more it can infer about your health, finances, or immigration status from data you never explicitly shared. Our new issue brief raises two urgent questions: Will agents act in users’ best interests when developer incentives diverge, and who is liable when they don’t? The authors argue that AI agent developers and deployers should be subject to a duty of loyalty and outline what a fiduciary framework would require from standards bodies, regulators, and Congress. Read more: https://lnkd.in/gitHV5tb

How is AI transforming empirical research? Stanford HAI's Stanford Causal Science Center and Stanford University Graduate School of Business are hosting the Empirical Methods in the Age of AI Conference on October 2-3, 2026. The conference brings together leading academics and industry practitioners to explore how AI is reshaping data collection, causal inference, econometrics, and research workflows. The program will feature keynote talks, panels, and discussions. Poster abstracts featuring new research, works in progress, and applied projects are welcome. Deadline: August 31. Learn more: https://lnkd.in/gWYvUezE

Large language models are trained on millions of authors — but converge on a single, homogeneous voice that flattens the variation between one human writer and another. A group of psychologists and computer scientists developed PsychAdapter to reintroduce human variation into AI-generated text by giving researchers control over personality, age, and mental health characteristics in language generation. The result opens new possibilities for therapy training, personalized education, and social science research: https://lnkd.in/gHjXDy6M