Miju Labs

← All people · Gray Swan AI

Head of CMU's Machine Learning Department, OpenAI board member chairing its Safety and Security Committee, and Gray Swan's chief scientist. Few people sit closer to how frontier labs think about adversarial testing.

high confidencePittsburghAt Gray Swan AI since 2023Updated 2026-09-19

Background

Stanford CS PhD and MIT postdoc; joined the CMU faculty in 2012 and now heads its Machine Learning Department. He has worked in industry as Chief Data Scientist at C3.ai and Chief Expert at the Bosch Center for AI. Joined OpenAI's board in August 2024 and chairs its Safety and Security Committee; also sits on Qualcomm's board and advises BNY.

Announced Gray Swan in July 2024 alongside Matt Fredrikson, Andy Zou and Dan Hendrycks.

What they run now

Research direction for Shade/Cygnal and the lab-facing red-teaming work (indirect prompt injection, agent security).

Career

  1. 2024 – presentBoard member; chair, Safety and Security Committee, OpenAI
  2. 2023 – presentCo-founder & Chief Scientist, Gray Swan AI
  3. 2012 – presentProfessor; Head of Machine Learning Department, Carnegie Mellon University
  4. —Chief Data Scientist, C3.ai
  5. —Chief Expert, Bosch Center for AI
  6. —Stanford University, PhD, Computer Science
  7. —MIT, Postdoc

On the record

AI-specific vulnerabilities · 2026-06

AI systems can be tricked much as people can, so securing them needs a different mindset from classic cybersecurity. source

Scale and robustness · 2026-06

Bigger models are not automatically more robust; safety requires explicit training. source

Agents and prompt injection · 2026-06

Agents are now smarter and actively look for prompt injections. source

Recent posts

26 posts archived · most engaged first, then the latest

Zico Kolter@zicokolter·
I'm teaching a new "Intro to Modern AI" course at CMU this Spring: modernaicourse.org. It's an early-undergrad course on how to build a chatbot from scratch (well, from PyTorch). The course name has bothered some people – "AI" usually means something much broader in academic contexts – but I think the time has come where the first thing that many students interested in AI should see is how the AI they are familiar with actually works (because it's really simple!) The more people who understand it the better. I'll be trying to put as much material as I can that we develop online (assignments + autograding, hopefully lecture videos), though as a first-time course there are also likely to be some bumps along the way. Hopefully it becomes a good resource over time, though. Feedback welcome.
502622.1K616KView on X ↗
Zico Kolter
Professor and Department Head at Carnegie Mellon University
•
I'm teaching a new "Intro to Modern AI" course this semester at CMU, and offering a free online version as well (starting January 26). Course info, and a link to enroll in the online version, is available at modernaicourse.org. In this course, students will build a simple LLM-based chatbot from scratch in PyTorch, without any pre-built layers or pre-trained models. The course name has been a bit controversial ("AI" usually means something much broader in academic contexts). But I think we're at the point where the _first_ thing that many students interested in AI should see is how the AI they actually use every day actually works. Plus, the beauty of these approaches is that they are incredibly simple, and it's really possible for a early undergraduate student to build it all. The more people that understand this, the better. In online version, students can watch all lecture videos and submit all the assignments (all of which are autograded in the full course anyway). There won't be quizzes/exams, and there's no credit offered via CMU; this is purely for educational purposes. The online course will begin two weeks after the CMU version, and lectures and assignments will all be released with the same two-week delay after the in-person version. This is a first-time course offering, so I'm sure there will be hiccups along the way, but I hope that many people will still find the content useful. I hope to see some of you there!
1,57333 comments • 112 reposts
Zico Kolter@zicokolter·
I am thrilled to have @paulfchristiano joining the OpenAI Foundation board and the Safety and Security Committee. Paul is a pioneer in AI Safety, Security, and Alignment, and I'm extremely grateful for the opportunity to work with him here.
4318315KView on X ↗
All 26 in the feed →

Interviews & talks

Connections

  • Matt Fredrikson — co-founder
  • Andy Zou — co-founder, former PhD student
  • Dan Hendrycks — named in Gray Swan launch announcement

Investments and boards

  • Board member, OpenAI (chairs Safety and Security Committee)
  • Board member, Qualcomm
  • Advisor, BNY

Who they amplify

Accounts whose posts Zico has reposted recently: Spencer Whitman, Rob Jenks, Gray Swan.

Why it matters here

Little direct relevance to creative data. His OpenAI board seat makes him the most lab-connected person on this list, but he is not a buyer of taste data.

How to reach

Not an early target. Academic route only (CMU ML department). Speculation: would engage only on research substance.

What we could not establish
  • Exact years at C3.ai and Bosch not verified here.
  • Wikipedia text describes his Gray Swan role as 'Senior Advisor' while the company and press say Chief Scientist; company title used.

Sources

  1. Zico Kolter bio
  2. Wikipedia: Zico Kolter
  3. Gray Swan About page
  4. Gray Swan: Series A announcement (2026-05-28)
  5. Latent Space: Red-Teaming after Mythos (2026-06-22)
  6. Zico Kolter on X: launching Gray Swan (2024-07)
Also at Gray Swan AI
high confidence

Matt Fredrikson

Co-founder & CEO

low confidence

Andy Zou

Co-founder (described as CTO in an undated CMU talk listing)

medium confidence

Rob Jenks

Chief Strategy Officer

low confidence

Spencer Whitman

Chief Product Officer