MATS empowers researchers to advance AI alignment, transparency, and security
Berkeley, CA
- Tweets 205
- Following 136
- Followers 5,847
- Likes 402
Our Research Managers are excellent coaches!
If you can push brilliant researchers toward their best, most impactful work, MATS wants to talk to you.
Apply for the Research Manager role by October 1st: matsprogram.org/careers/rese…
1
9
981
MATS Research retweeted
I've added the reading list for our MATS stream (co-led with @LRudL_) to my website. It's a good starting place for AI strategy & governance work, especially if you're focused on distributing power/preventing power concentration.
A 🧵 on some of my favorite recommendations:
6
32
279
20,083
MATS Research retweeted
🤖👿 New paper on Misalignment via In-Context Persona Induction! 🤖👿
We show that one can induce personas in-context, on frontier LLMs (~5+ months old models, things move quick!), which can dramatically degrade their alignment!
Link: arxiv.org/pdf/2609.06851
5
14
93
8,460
MATS Research retweeted
This summer I was in Berkeley as part of @MATSprogram working on technical governance.
At the end of the program I gave a short symposium talk. You can check it out here: piped.video/watch?v=G8d2uDdq…
1
1
4
740
MATS Research retweeted
I'm building a new org to try to understand how close we are to RSI and explosive economic growth by systematically estimating the elasticities of substitution, as part of the @MATSprogram.
docs.google.com/document/d/1…
23
30
338
27,771
MATS Research retweeted
New paper: World Modeling in Transformers
with @matthieu_queloz @andrenfreitas @MATSprogram
Can AI models recover world models just from data? We find a faithful internal map and more inside of taxiGPT! We trace failures to how that the map is stored in superposition. 🧵(1/11)
7
62
479
67,438
9/ Researchers with a serious track record in AI safety or adjacent fields (ML, policy, or governance) who want to advance research that reduces catastrophic risk are encouraged to apply.
Questions? Email residency@matsprogram.org.
🔗 Apply here: matsprogram.org/residency/ap…
1
29
1,549
8/ Every Resident leaves with a substantial body of reviewable work, 4 to 8 weeks of transition support, & ongoing access to the MATS community.
1
20
1,515
7/ Applications for the 2027 Winter intake:
→ Opens 25 August 2026
→ Closes Oct 31, 2026
→ Application outcomes sent late Nov 2026
→ Program begins in early 2027
Up to 30 Residents accepted per year across 3 intake windows.
1
21
1,502
6/ What Residents receive:
→ $155,000–$285,000/yr compensation
→ Full health, dental, vision, & life insurance
→ Relocation support up to $25K + visa assistance
→ Research compute with no preset cap
→ Dedicated conference budget & catered meals
1
27
1,586
5/ Anchor: a staff researcher path.
For applicants who want a 2-year staff researcher role rather than a PhD or postdoc.
An alternative pathway for engineers, evaluators, and technical or policy practitioners whose work is fundamentally research.
1
27
1,708
4/ Bridge: for experienced professionals moving into AI safety.
For academics, researchers, and industry people pivoting from ML, policy, or governance. Strengthen your portfolio and produce work that demonstrates your expertise.
Sabbaticals of up to 2 years supported.
1
30
1,921
3/ Incubation: launching a new research paradigm or organization
For scoping original research agendas or founding new AI safety research orgs.
We help you expand your network, build credibility with collaborators and funders, and access what you need to launch.
2
30
2,352
2/ Having supported 630+ safety researchers who have co-authored 215+ papers with 17,000+ citations, we built the Residency for long-term work that doesn’t fit standard academic or industry settings.
We offer 3 specialized tracks. ↓
1
38
3,568
1/ MATS is launching the Residency program!
A new path for experienced researchers working on AI safety, or moving into it.
📍 Berkeley / London / Washington, DC
💰 $155,000–$285,000/yr + compute, no preset cap
📅 6–24 months
Apply by October 31 AoE ↓
matsprogram.org/residency?ut…
11
60
528
461,759
Reducing risks from powerful AI is the world’s most urgent & talent-constrained challenges.
We aim to train the next generation of AI safety researchers & founders.
Know someone who’d be a strong fit? Share this thread.
🔗 matsprogram.org/apply
🚨 September 6, 2026 AoE
1
7
56
10,319
🌟 Featured tracks:
→ Policy & Governance: how advanced AI is governed + how it should be
→ Biosecurity: risks at the AI × bio intersection
→ Founding & Field-Building: launching new AI safety initiatives
Explore all seven tracks & find your fit ↓
1
23
9,277
🎯 MATS has seven tracks:
→ Empirical: ML experiments on advanced AI safety
→ Theory: foundations of alignment, agency & safe reasoning
→ Strategy & Forecasting: understanding how advanced AI development may unfold
→ Systems Security: securing advanced AI infrastructure
1
28
8,183
Our track record:
→ 631 alumni, 100+ mentors
→ 215+ publications, 17,000+ citations
→ 75% of pre-2026 alumni work in AI safety/security
→ 30+ alumni-founded initiatives
1
29
8,914