MATS empowers researchers to advance AI alignment, transparency, and security
Joined November 2023
- Tweets 145
- Following 136
- Followers 3,957
- Likes 316
3 Photos and videos
Pinned Tweet
May 13
1/ π¨ MATS Autumn 2026 applications are now open.
10-week fully-funded fellowship for aspiring AI alignment, security & governance researchers and field-builders.
π Berkeley London
π
Sep 28 β Dec 4, 2026
π° $5000/month stipend $8,000/month compute
Apply by June 7 AoE β
10
96
747
3,264,535
Jun 8
π¨ Applications for MATS Autumn 2026 close tonight (June 7 AoE)!
Spend 10 weeks fully funded working with mentors from Anthropic, DeepMind, OpenAI, Redwood Research, SecureBio, and more.
New this cohort:
𧬠Biosecurity
π Founding & Field-Building
Apply now: matsprogram.org/apply
6
11
153
57,178
MATS Research retweeted
Jun 5
Applications for @MATSprogram close in two days! Learn more about our stream and apply today (link in replies)
May 28
If you want to research interventions to gradual disempowerment or the intelligence curse, @LRudL_ and I are mentoring a @MATSprogram stream this autumn.
Many people have asked me βwhatβs the plan to make this go well?β Right now, thereβs not one. You should help fix that.
π§΅
1
4
34
4,706
MATS Research retweeted
Jun 4
Trained monitors can be strong low-cost alternatives to prompted frontier models for black-box scheming detection.
Our fine-tuned open-weight monitors detect scheming/sabotage in agent trajectories better than small prompted models and are on the cost-performance frontier. (1/n)
ALT Cost-performance frontier for action-only scheming monitors. Our trained monitors provide a low-cost option for low-FPR monitoring. Across six OOD datasets, our trained monitors outperform smaller prompted frontier monitors for under $5 per 1,000 evaluations, while larger frontier monitors are substantially more expensive.
1
7
29
6,223
MATS Research retweeted
Jun 4
New research from @japhba and I!
Activation Oracles are a pretty cool interpretability tool. They answer natural questions about activations, but they suffer from vagueness and hallucinations. Can AO training be improved?
Turns out: Yes! We identify four fixes that make AOs substantially more useful!
12
24
224
30,862
MATS Research retweeted
Apply by June 7 to work with MATS in the Fall.
In the MATS stream that I coordinate, we will work on impact-oriented technical AI governance research, potentially including research on open-weight models, AI safeguards, AI incidents, technically rigorous AI policy, etc.
May 13
1/ π¨ MATS Autumn 2026 applications are now open.
10-week fully-funded fellowship for aspiring AI alignment, security & governance researchers and field-builders.
π Berkeley London
π
Sep 28 β Dec 4, 2026
π° $5000/month stipend $8,000/month compute
Apply by June 7 AoE β
1
7
118
8,874
MATS Research retweeted
Excited to serve as a mentor on the @MATSprogram founding & field-building track this autumn. If you're interested in moving into AI security and alignment, critical cyber security, or biosecurity, you should consider applying. Deadline is June 7th.
May 29
π¨ New for MATS Autumn 2026: the Founding & Field-Building track.
A fully-funded track for founders, field-builders and amplifiers ready to launch and scale new AI safety initiatives.
Apply by June 7 AoE β
1
1
663
MATS Research retweeted
May 31
Apply by June 7th to work with me on the institutional stack for the post-AGI world: moral reasoning, bargaining, structured transparency, trust infrastructure, and resilience capabilities.
May 29
π¨ New for MATS Autumn 2026: the Founding & Field-Building track.
A fully-funded track for founders, field-builders and amplifiers ready to launch and scale new AI safety initiatives.
Apply by June 7 AoE β
10
115
11,483
MATS Research retweeted
Jun 1
This trackβand MATS more broadlyβis a great way to develop an idea for a new AI safety org to the point where it's ready for CG funding. I bet we'll end up making some seven- or eight-figure grants to new orgs that come out of MATS this year.
May 29
π¨ New for MATS Autumn 2026: the Founding & Field-Building track.
A fully-funded track for founders, field-builders and amplifiers ready to launch and scale new AI safety initiatives.
Apply by June 7 AoE β
4
99
7,493
MATS Research retweeted
Jun 2
Geoff Ralston @gralston and I are teaming up to mentor a biosecurity stream for @MATSprogram this fall. Check out the other great streams!
May 29
Streams led by:
Β· @dewierwan_ (BlueDot)
Β· @MikeMcCormick_ & @nickraushenbush (@HalcyonFutures )
Β· @gralston & @incredutility (Bio Action Plan)
Β· @LisaThiergart & @luiscosio (@SL5TaskForce)
Β· @AmmannNora (ARIA)
Β· @RosieCampbell (@eleosai )
Β· @degerturann (@metaculus)
3
11
1,464
MATS Research retweeted
Iβm mentoring Autumn 2026 @MATSprogram Fellows interested in doing AI welfare research.
The application deadline is this Sunday (6/7). More info in this thread:
May 13
1/ π¨ MATS Autumn 2026 applications are now open.
10-week fully-funded fellowship for aspiring AI alignment, security & governance researchers and field-builders.
π Berkeley London
π
Sep 28 β Dec 4, 2026
π° $5000/month stipend $8,000/month compute
Apply by June 7 AoE β
2
7
102
8,802
MATS Research retweeted
Jun 2
You should really follow @rocketalignment.
You donβt wanna miss gems like this dive into exciting research by @MATSprogram.
1
4
15
951
MATS Research retweeted
Jun 2
I am going to be mentoring for a new MATS track focused on founders and amplifiers!
Many fellowships focus on research, but there's so much to be done beyond that. Come found orgs, build infra, run events, and help us scale up the field of AI welfare.
Apply by June 7 matsprogram.org/apply
May 29
π¨ New for MATS Autumn 2026: the Founding & Field-Building track.
A fully-funded track for founders, field-builders and amplifiers ready to launch and scale new AI safety initiatives.
Apply by June 7 AoE β
6
101
9,507
MATS Research retweeted
May 29
MATS Autumn applications due June 7!
Pitch: Come work with me and Alex Cloud in Team Shard! We have fun, consistently make real alignment progress (we pioneered steering vectors in 2023!), and help scholars tap into their latent abilities.
3
9
151
10,600
May 29
π¨ New for MATS Autumn 2026: the Founding & Field-Building track.
A fully-funded track for founders, field-builders and amplifiers ready to launch and scale new AI safety initiatives.
Apply by June 7 AoE β
2
26
262
48,845
May 29
The essentials:
π
10 weeks Β· Sep 28 β Dec 4, 2026
π° $5k/mo stipend $8k/mo compute & research budget
π Berkeley / London / remote
π Housing, meals, travel, J-1 visa
β 6β12 month extension opportunity
1
12
1,822
MATS Research retweeted
May 27
π¨ New research work with @CHAI_Berkeley!
We provide the first multi-domain benchmark evaluating safety monitors for OOD misalignment detection by intentionally restricting the training dataset.
Special thanks to folks @MATSprogram and @haizelabs for providing valuable feedback and compute.
May 27
We've seen AI models deceive, gaslight, and drive users to psychosisβsafety issues that labs didn't anticipate until they caused real harm. We built the first benchmark of these unknown unknown alignment failures and found that OOD detection can help prevent them. π§΅
2
7
1,092