MATS Summer 2026 cohort launches this week — 120 fellows, 100 mentors, methodological tracks spanning post-SAE interpretability, behavioral analysis, and capability-eval frameworks
The MATS (ML Alignment & Theory Scholars) Summer 2026 cohort begins this week with 120 fellows and 100 mentors — roughly double the 2024 program size. Tracks span mechanistic interpretability, behavioral analysis, training-data influence, capability-evaluation frameworks, and theoretical alignment. The cohort is the dominant entry-tier pipeline for new alignment researchers, and its expansion comes alongside funding commitments from Anthropic, OpenAI, DeepMind, and Open Philanthropy.
The substantive piece is methodological diversity at launch. DeepMind's just-announced SAE deprioritization hits the MATS curriculum at exactly the moment the field's dominant methodology is being questioned. The cohort's 5-track structure is well-positioned for this — fellows can move toward behavioral analysis, capability evals, or training-data influence tracks rather than locking into a single methodology now under contestation.
The pipeline math is that 120 fellows generate roughly 60-80 publications across the August demo period, which then feeds the lab hiring pipelines for the September-October cycle. Combined with Anthropic's $150M Claude Corps program at 1,000 fellows, the AI-safety-trained labor force entering the market in 2027 is structurally larger than any prior year.
MATS Program — MATS Summer 2026 → · Zylos Research — AI Safety, Alignment, and Interpretability in 2026 → · ArXiv — Mechanistic Interpretability for AI Safety — A Review →