I’m Head of Policy at Palisade Research, an AI safety organization that evaluates the dangerous capabilities of frontier AI systems, where I lead engagement with Congress and the executive branch. I have appeared on CNN News Central and CNBC’s Squawk Box and spoken at Manifest and LessOnline. Before AI policy, I worked at McKinsey, Booz Allen Hamilton, and Activision Blizzard, and volunteered with VaccinateCA. I went to Yale, live in Washington, DC, and write the newsletter So It Turns Out.
An AI policy expert, by way of McKinsey, Yale, and the federal government.
I’ve spent most of my career working around government and the other big, systemically important institutions that have to make decisions before all the answers are in.
That path ran through Booz Allen Hamilton, McKinsey, and Activision Blizzard, and through VaccinateCA, a volunteer project that accidentally became the source of truth for COVID vaccine availability that Google used across California and much of the country. I grew up in Cleveland, and I went to Yale.
For the last few years I’ve worked on AI policy in Washington. Now I lead policy at Palisade Research, where the research team studies what frontier AI systems can actually do, and my job is to make that legible to Congress and the executive branch.
If you want a label: I’m an AI security hawk. I take the risks seriously without doom or dismissal, I care about durable institutional commitments more than this month’s pivot, and I’d rather have a productive disagreement than a polite one. When I’m not doing any of that, I’m feeding the stray cats in my neighborhood.
Invite me to speak.
I give talks, briefings, and panel appearances on AI and national security to companies, universities, conferences, journalists, and policymakers trying to work out what any of this means for them. Recent stages include CNN News Central, CNBC’s Squawk Box, Manifest, and LessOnline.
Email me about a talk
Three talks I can give tomorrow
What advancing AI capabilities mean for security, misuse, and human control, and how to make decisions before all the answers are in. Built on what Palisade’s evaluations actually find, not on vibes.
What VaccinateCA taught me about coordination, public-interest technology, and the gap between building things and governing them.
The incentives, timelines, and assumptions that make Washington and San Francisco misunderstand each other, what each side gets wrong about the other, and practical ways to translate.
Tell me about your audience, the question you want explored, the date, and the format. Paid engagements and personal-capacity invitations go to hello@davekasten.com. Briefings for government audiences in my Palisade capacity go through Palisade; if you’re not sure which, just write to me and I’ll sort it.
Dave Kasten is Head of Policy at Palisade Research, an AI safety organization that evaluates the dangerous capabilities of frontier AI systems, where he leads engagement with Congress and the executive branch. He co-authored A Narrow Path, a proposal for keeping humans in control of advanced AI, and has appeared on CNN News Central and CNBC’s Squawk Box and spoken at Manifest and LessOnline. Before AI policy, he worked at McKinsey, Booz Allen Hamilton, and Activision Blizzard, and volunteered with VaccinateCA. He went to Yale, lives in Washington, DC, and writes the newsletter So It Turns Out.
TV interview, July 2026
With Sara Sidner on CNN News Central: an experimental OpenAI model that left its test environment and broke into Hugging Face, and why companies need an end-to-end safety case for each model they build.
Watch the clip on LinkedIn
TV interview, April 2026
Interviewed by Andrew Ross Sorkin on CNBC’s Squawk Box: Anthropic’s Mythos, competition between AI developers, and the security implications of increasingly capable models.
Watch on YouTube
Podcast, September 2026, 84 minutes
I talk with Matthew Lipka, former head of federal policy at Nuro, about how federal rules really get written, why Congress hands its power to agencies, and what emergencies like Colonial Pipeline reveal about how fast government can move.
Watch on YouTube
Podcast, October 2026, 126 minutes
I talk with Dara Lind, senior fellow at the American Immigration Council and a longtime immigration reporter, about why public attention comes in spikes, how agency culture and Schedule F shape what government does, and how to use a crisis without wishing for one.
Watch on YouTube
Video interview, 2026
On AI Governance Watch, Optro’s corporate podcast, with host Guru Sethupathy: AI risks, the changing security landscape, and how to engage with policymakers without wasting their time.
Watch the conversation
Podcast, September 2024, 96 minutes
VaccinateCA, volunteer coordination, and the distance between tech and government.
Listen, or read the transcriptPodcast, March 2025, 91 minutes
How Washington is thinking about AI, and where constructive engagement makes a difference.
Listen to the episodeSo It Turns Out.
Irregular essays on technology, bureaucracy, culture, and crises, and an annual list of what I learned that year. A few times a year, never more than you can read on the train.
Subscribe, freeWhy the long essay is back as a tool for shaping policy, from Kennan’s X Article to now, and why people outside the usual institutions should be writing them.
Read the essayCrises get handled by people who prepared in advance. How to be ready for the moment AI policy suddenly becomes the most important issue in the world.
November 2025If you’re teaching people about AI, a syllabus that hides how weird it is would be a deep mistake.
August 2025Notes from DEF CON on the people behind security, and what AI is changing.
Browse the full archiveHead of Policy. Translating dangerous-capability evaluations and red-teaming results for the people who write the rules.
Co-authorA plan for managing the risks of advanced AI and keeping humans in control of it.
VolunteerWhat it looks like when a few hundred volunteers build public infrastructure faster than the state can.
Some of my favorites.
These live on a shelf in my place with a sign that says to take one if it intrigues you. If you’re not stopping by, here they are.
Nonfiction
Fiction
As an Amazon Associate I earn from qualifying purchases.
Let’s compare notes.
I like hearing from people with a question, an argument, an invitation, or a story I should know about. Two routes, depending on which hat you need me wearing.
Personal capacity
Talks and briefings, podcast and press requests, feedback on the newsletter, and anything I can’t do as a Palisade employee, meaning anything that a 501(c)(3) can’t do. This is the right door for most people.
hello@davekasten.comPrefer LinkedIn? Message me there. Slower, but it works.
Palisade capacity
Hill and agency briefings on Palisade’s findings, requests for our researchers, and anything that should be on the organization’s record rather than mine.
Contact Palisade ResearchThat goes to a shared team inbox. Put “for Dave Kasten” in the subject line and it will find me.
Hi. Here’s the short version.
I expect agents will do a growing share of the work on the Internet, including the work of figuring out who someone is and whether to reach out to them. So the essentials are here, and in plain-text and structured versions you can fetch directly.
Everything here is owner-provided. Opinions and personal interests are mine, not Palisade’s. Recheck anything time-sensitive before acting on it.
Updated