AI Safety Training · Week 46, 2026
This is a weekly newsletter that lists newly announced training programs addressing existential risk from AI. Visit AISafety.com/training for the full directory of upcoming programs.
New training programs
• Iliad Fellowship: February 2027
Fellowship · Bay Area, USA & London, UK · 3 months · Starts 1 Feb 2027
Expenses covered · Full-time · Entry bar: high · Focus: technical · Apply by 16 Dec 2026
Mentored research in applied mathematics for alignment, preceded by a course covering the field. For PhD students, postdocs, and researchers of equivalent experience. About half of fellows go on to a research extension.
• ARBOx5 (Alignment Research Bootcamp Oxford)
Bootcamp · Oxford, UK · 12 days · Starts 4 Jan 2027
Expenses covered · Full-time · Entry bar: mid · Focus: technical · Apply by 14 Nov 2026
Run by the Oxford AI Safety Initiative (OAISI) with a curriculum adapted from ARENA. Covers ML fundamentals, building GPT-2 small, mechanistic interpretability, RL, and evals – ending with a short capstone project.
• Talos Fellowship: Spring 2027
Fellowship · Online & Brussels, Belgium · 7 weeks · Starts 24 Feb 2027
Expenses covered · Part-time · Entry bar: mid · Focus: governance · Apply by 6 Dec 2026
Program by Talos Network for recent graduates and early-career professionals who want to work in European AI governance. Includes a weekly seminar, 1-1s, a published policy piece, a summit, and an optional placement.
• Policy Leaders Programme: Spring 2027
Fellowship · Online & Brussels, Belgium · 8 weeks · Starts 23 Feb 2027
Expenses covered · Part-time · Entry bar: high · Focus: governance · Apply by 6 Dec 2026
Program by Talos Network helping experienced professionals in policy, law, defense, and similar fields apply their expertise to European AI governance. Includes discussion calls, 1-1s, a summit, and optional placement.
• AI Evaluation Programme: 2027
Course · Online · 5 months · Starts 1 Feb 2027
No stipend · Part-time · Entry bar: mid · Focus: technical · Apply by 11 Dec 2026
For graduate students, early-career researchers, and professionals. Modules include adversarial evaluations and red teaming, mechanistic interpretability, and misuse and extreme risk – alongside a team capstone study.
• AI Safety Collective at Irvine (AISCI) Research Projects
Fellowship · Online & Irvine, USA · 3 weeks · Starts 20 Nov 2026
No stipend · Part-time · Entry bar: mid · Focus: technical · Apply by 17 Nov 2026
Teams of 1–3 work on open questions in technical AI safety, each mentored by a board member. Projects cover monitoring AI agents for attacks, evaluation awareness, and jailbreak defenses – ending with a blog post.
• SAGE: S-risk and AI Governance Exchange
Course · Online · 6 weeks · Starts 11 Jan 2027
No stipend · Part-time · Entry bar: high · Focus: governance · Apply by 22 Nov 2026
Working group by the Center for Reducing Suffering (CRS) on how AI governance could reduce s-risk (risk of astronomical suffering). People from both fields rank open questions and identify potential interventions.
• AGI Strategy Course
Course · Online · 5 weeks · Starts 11 Jan 2027
No stipend · Part-time · Entry bar: low · Focus: general
Program in Spanish by Seguridad de la IA (iaS) using BlueDot Impact's AGI Strategy curriculum. Covers how AI is developing, the risks and opportunities of different outcomes, and how to steer AI toward better scenarios.
• 1st Benevolent AI Workshop
Immersive workshop · İzmit, Turkey · 2 days · Starts 5 Dec 2026
No stipend · Full-time · Entry bar: low · Focus: general
Program by Poedat introducing AI safety and governance in Turkish. Covers alignment, reward hacking, catastrophe scenarios, and control of advanced AI. Includes talks, paper discussions, and a career planning workshop.