MATS Summer 2026

Neel takes a pragmatic approach to interpretability: identify what stands between where we are now and where we want to be by AGI, and then focus on the subset of resulting research problems that can be tractably studied on today's models. This can look like diving deep into the internals of the model, or simpler black box methods like reading and carefully intervening on the chain of thought - whatever is the right tool for the job. This could look like studying how to detect deception, understanding why a model took a seemingly concerning action, or fixing weak points in other areas of safety, e.g. using interpretability to stop models realising they are being tested. You can learn more about Neel's approach in this podcast.

He has spent far too much time having MATS scholars, and has worked with ~60 so far - he’s excited to take on even more!

We are interested in mentoring projects in AI forecasting and governance. This work would build on the AI 2027 report to either do more scenario forecasting or explore how to positively affect key decision points, informed by our scenario.

Agent Foundations research focused on clarifying conditions under which humans can justifiably trust artificial intelligence systems.

Alignment is solved for models in the current paradigm. This shifts the threat model to good old human conflict, so I'm excited about coordination tech (AI cooperation, datacenter workload verification). For aligning future models, we have to forecast what future AGIs will look like and solve issues before they come up. I’m excited about models that maintain their goodness under self-directed learning and can align their successor.

This stream focuses on empirical AI control research, including defending against AI-driven data poisoning, evaluating and attacking chain-of-thought monitorability, and related monitoring/red-teaming projects. It is well-suited to applicants already interested in AI safety with solid Python skills, and ideally prior research or familiarity with control literature/tools (e.g. Inspect/ControlArena). 

Building realistic defensive cybersecurity benchmarks. Asymmetric Security responds to real cyber incidents and therefore holds data not available in the public domain. We would like to work with MATS scholars to build realistic benchmarks grounded in these real cyber incidents. 

The Alignment Research Center is a small non-profit research group based in Berkeley, California, that is working on a systematic and theoretically grounded approach to mechanistically explaining neural network behavior. We are interested in scholars with a strong math background and mathematical maturity. If you'd be excited to work on the research direction described in this blog post – then we'd encourage you to apply!

This coalition of mentors make up the “megastream”. This stream spans a range of empirical research areas in AI safety on LLMs, including AI control, scalable oversight, model organisms, model internals, model welfare, security, and more. You’ll be pitched, and have the option to pitch, a variety of safety research projects, and then be matched to projects and mentors based on your interests/preferences on research and what you’d like to get out of MATS. Scholars in this stream frequently receive funding and continued mentorship after MATS to complete their research project, usually leading to a (co-)first author paper. People in this stream often end up in long-term homes for safety research after MATS (e.g. Anthropic, Redwood Research, OpenAI).

​

Megastream mentors share an application, tend to collaborate and co-mentor projects together, and generally share infrastructure to streamline the scholar experience. By applying to this stream, you are being considered for all of the megastream mentors. In the application process, you can indicate particular mentors you are interested in working with.

Arthur Conmy's MATS Stream focuses on evaluating interpretability techniques on current and future AI Safety problems.

​

This can involve creating new safety techniques, as well as creating benchmarks and measuring performance against baseline techniques.

In the face of disaster, I suspect the government will be forced to play insurer of last resort, whether for a particular lab, or society at large. (I'm not the only to suspect this – see e.g. here). Designed well, I believe a federal insurance backstop could internalize catastrophic negative externalities; designed poorly, it will simply be a subsidy for AI companies. I want to design the good version, so we have it ready.

​

I encourage people with mechanism design (a.k.a. reverse game theory) expertise to apply, but don't be deterred if you don't have this expertise.

This stream focuses on representations that underlie how language models generalize, for example representations of personas, goals, or training data components.

We study applications of singular learning theory (SLT) to AI safety, with a focus on interpretability and alignment. Ideal candidates come from a strong technical background in mathematics, physics, computer science, or biology, and aren't afraid to get their hands dirty with ML experiments. We don't expect you to have deep expertise in SLT, but a shallow familiarity will help. 

I have two broad areas.

​

Security:

I am interested in building demonstrations for hacking real-world AI deployments to show that they are not secure. The goal is to force companies to invest in alignment techniques that can solve the underlying security issues.

​

Benchmarks:

I am interested in building benchmarks to determine how generalizable modern LLM techniques actually are, now that we are no longer in the pre-training scaling era.

This stream will focus on monitoring, stress-testing safety methods, and evals, with a focus on risks from scheming AIs. Examples include (black-box) AI control techniques, white-box monitors (probes etc.), chain-of-thought monitoring/faithfulness, building evaluation environments, and stress-testing mitigations.

This stream will work on gathering and analyzing data in order to shed light on the driving forces behind AI and monitor its impacts.

AI macrostrategy: strategic questions about how the transition to advanced AI will happen, and what we can do now to prepare for it.

​

Topics of interest include better futures, power concentration, takeoff speeds, deals with AIs, space governance, and acausal trade.

In this project, we will explore GPU side-channel attacks to extract information about model usage. A simple example is to observe (via radio, power fluctuations, acoustics, etc.) which experts were used in each forward pass of an MOE model, then use those observations to guess which tokens were produced.

I'm interested in mentoring projects related to reward hacking and monitoring (agentic) models that produces long and complex trajectories. Scholar will have freedom to propose projects within this scope. Expect 30-60min 1-1 time on zoom.

Janet Egan will mentor scholars working on policy-relevant questions at the intersection of AI compute, geopolitics, and infrastructure. Potential projects include analyzing remote access to AI chips (e.g., via cloud providers in China), mapping and interpreting the global buildout of AI data centers and energy infrastructure, and developing politically informed strategies for US–China cooperation on AI risk. The mentee will lead their research project with weekly guidance, feedback, and optional career and policy insights.

Implementing SL4/5 and searching for differentially defense-favored security tools.

This stream will pursue research on securing and hardening AI systems through rigorous testing, provable defenses, and formal specification, including improving benchmarks for agentic security, scaling mathematically-grounded robustness techniques like randomized smoothing and Lipschitz-constrained training, and developing formal methods for specifying safe agent behaviors.

We are excited to supervise projects that fall within the two following categories:

1. Studying the causes, implications, and mitigations of [instances of] situational awareness;
2. Contributing directly to LawZero's Scientist AI. 

​

For 1., we are particularly interested in:

• Evaluation / monitorability awareness;
• Self-awareness, in an introspective sense.

​

For 2., we are especially interested in:

• Testing if  "truth-ification" (a process that, given a corpus of text, augments it so as to make sources of information explicit) allows language models to generalize better;
• Developing amortized inference methods to estimate the uncertainty of a predictor (such as an autoregressive model). 

Lee's stream will focus primarily on improving mechanistic interpretability methods for reverse-engineering neural networks.  

This stream will work on projects that empirically assess national security threats of AI misuse (CBRN terrorism and cyberattacks) and improve dangerous capability evaluations. Threat modeling applicants should have a skeptical mindset, enjoy case study work, and be strong written communicators. Eval applicants should be able and excited to help demonstrate concepts like sandbagging elicitation gaps in an AI misuse context.

Priority directions:

• Risks from automating AI research
• Automating safety and alignment research
• AGI privacy
• Measuring long-horizon agentic capabilities
• New alignment methods
• Science of post-training

We will continue working on black-box monitors for scheming in complex agentic settings, building on the success of the previous stream.

​

See here for details.

AI control focussed stream, probably running in-person in London.

Escalation risks from state perceptions of AI capability, AI-enabled targeting, AI-enabled decision manipulation, and the impact of AI integration into nuclear command and control. 

This stream focuses on AI policy, especially technical governance topics. Tentative project options include: technical projects for verifying AI treaties, metascience for AI safety and governance, and proposals for tracking AI-caused job loss. Scholars can also propose their own projects.

This stream will focus on the science and development of model evaluations, especially monitorability and alignment evals.

Research papers (technical governance or ML) related to evaluating and mitigating dangerous AI capabilities, with a focus on what's actionable and relevant for AGI companies

This stream will focus on projects to better understand the capabilities of the model on dangerous capabilities specially more related to security. 

​

Also finding better ways to evaluate the safety and robustness of the models.

I'm interested in empirical projects that improve our ability to evaluate model capabilities or enable to understand or evaluate model monitorability. An ideal project culminates in a research output (conference/Arxiv paper or research blogpost with artifacts).

Making society safe from AI doesn't just mean making safe AI: we're figuring out how to uplift human collective intelligence, manage a highly multiagent world, improve foresight and institutional competence, ideally learning how to make best positive use of frontier AI systems as we go. FLF has a small, sharp team of researchers with a wide network, and we're looking to nurture new and missing approaches to minimising large-scale risks while steering to a flourishing future.

Projects in this stream will be on AI welfare and moral status; more specifically, on what it takes to be a moral patient and how we can determine whether AI systems meet the conditions. I'm looking for applicants who have ideas about these topics and are motivated to explore them in more detail.

In this stream we will explore extensions and implications of our discovery that neural networks pretrained on next-token prediction represent belief-state geometry in their activations. We will build on this fundamental theory of neural network representations in order to discover what AI systems are thinking, and understand their emergent behaviors.

Peter Henderson’s stream focuses on developing safe, aligned AI agents, with projects on scalable oversight rules informed by law and game theory, safe long-horizon exploration, and measuring “jagged” capability/safety frontiers. Scholars will join an independently driven, engineering-heavy research environment, collaborating with other MATS scholars and PhD students, with weekly 1:1s and active async mentorship.

The Redwood Research stream is looking for fast empirical iterators and strategists to work on control research.

My MATS fellows will do philosophical thinking about multi-agent intelligence and how agents change their values. This will likely involve trying to explore and synthesize ideas from game theory, signaling theory, reinforcement learning, and other related domains.

Roger Grosse’s stream investigates how to improve influence functions and other training data attribution methods, and uses these tools to study alignment-related phenomena such as out-of-context reasoning and emergent misalignment. The ideal scholar has experience with LLM internals, strong statistics/applied math skills (especially numerical linear algebra), and can independently drive research from literature review through experimentation and analysis. Roger provides shovel-ready projects while giving exceptional scholars freedom to pursue their own ideas, and is open to scholars collaborating with others.

International coordination to reduce frontier AI risks, with a focus on China and the West. 

We build scalable technology for AI understanding and oversight.

The stream will advance empirical methodologies for third-party AI safety evaluations. Example research topics include chain-of-thought monitorability, the secret loyalties research agenda, and automatic auditing (eg, with Anthropic’s Parallel Exploration Tool for Risky Interactions).

The stream will focus on conceptual, empirical, and theoretical work on scalable oversight and control. This includes but is not limited to creating model organisms for specific failure modes, designing training procedures against them, and making progress on subproblems involved in safety cases.

I (Cas) work on a range of projects from technical safeguards to technical governance. This stream follows an academic collaboration model and will work will likely focus on technical topics in AI governance. 

​

In the shard theory stream, we create qualitatively new methods and fields of inquiry, from steering vectors to gradient routing to unsupervised capability elicitation to robust unlearning. If you're theory-minded, maybe you'll help us formalize shard theory itself.

I mostly interested in AI control and scalable oversight. I'm excited to work with scholars interested in empirical projects building and evaluating control measures and oversight techniques for LLM agents, especially those based on chain of thought monitoring. I'm also interested in the science of chain of thought monitorability, misalignment and control. An ideal project ends with a paper submitted to NeurIPS/ICML/ICLR.

This stream is for the UK AISI Red-team. The team focuses on stress-testing mitigations for AI risk, including misuse safeguards, control techniques and model alignment red-teaming. We plan to work on projects building and improving methods for performing these kinds of evaluations and methods.

Conceptual research on deceptive alignment, designing scheming propensity evaluations and honeypots. The stream will run in person in London, with scholars working together in team(s).