MATS Winter 2027 / Summer 2027 Mentorship

The MATS Program supports established AI alignment, interpretability, governance, and security researchers in mentoring talented researchers. Mentors guide a small stream of fellows through a substantive ten-week research project, supported throughout by dedicated research management which handles scoping, coordination, and day-to-day oversight, leaving the mentor free to attend to intellectual direction.

Mentoring with MATS provides:

  • The opportunity to work with and influence top-quality research talent

  • Dedicated research management support for your stream

  • Compute funding of $2,000 per week per fellow

  • The option to continue with fellows into a 6 to 12 month extension phase for longer-horizon work

Mentorship is compatible with full-time appointments elsewhere; time commitments are agreed individually.


Impact

527 fellows have now passed through MATS, and the program has produced 160+ co-authored papers attracting 7,800+ citations. Mentors who have run streams with us have seen their fellows go on to roles at Anthropic, Google DeepMind, OpenAI, the UK AI Security Institute, and more; around 10% of alumni have gone on to co-found AI safety organisations of their own, including Apollo Research and Timaeus. The research direction set during the fellowship frequently shapes the trajectory a fellow continues on.