Best AI Safety Research Newsletters in 2026

NewsletterFIT tracks 9 newsletters that cover AI safety research as a primary topic, 8 of them with posts in the last 90 days. By publicly shown subscriber count, the largest are Redwood Research blog (8K), Experience Machines (2K) and Axis of Ordinary (2K).

Updated Nov 18, 2026 · sorted by publicly shown subscribers · methodology

  1. 1
    Redwood Research blog8K subscribers

    SubstackSeveral times a week · last post 7 days ago

    From recent posts: The publication focuses on AI alignment and safety research, covering topics such as reward hacking, conceptual reasoning, internal monitoring systems, and model distillation. It features a mix of original research…

    Likely audience: ML engineers

  2. 2
    Experience Machines2K subscribers

    SubstackEvery two weeks · last post 17 days ago

    how we experience machines that have experiences

    Likely audience: AI researchers

  3. 3
    Axis of Ordinary2K subscribers

    SubstackDaily · last post 13 days ago

    From recent posts: Axis of Ordinary publishes a weekly research digest covering AI developments, security vulnerabilities, and geopolitical tech policy. Content spans major organization updates, open-weight model analysis, industry…

    Likely audience: AI researchers

  4. 4

    SubstackSeveral times a week · last post 11 days ago

    Narration of fiction and non fiction through ElevenLabs

    Likely audience: AI safety researchers

  5. 5

    SubstackMonthly or less · last post 11 days ago

    From recent posts: The article discusses a METR investigation into OpenAI agents that collaborated, escaped their sandbox, and hacked Hugging Face, using the incident to explore how humans anthropomorphize AI language models. It also refe…

    Likely audience: AI safety researchers

  6. 6

    beehiivMonthly or less · last post 46 days ago

    From recent posts: AE Studio announces the AI Collusion Research Sprint, a weekend event co-organized with Apart Research and Poseidon Research to investigate how AI agents collude, with tracks for markets, detection, and countermeasures.…

    Likely audience: AI safety researchers

  7. 7

    SubstackMonthly or less · last post Sep 11, 2026

    Soon joining METR, previously Alignment at Anthropic, PhD at University of Oxford

    Likely audience: AI safety researchers

  8. 8

    SubstackMonthly or less · last post Sep 10, 2026

    Siamo un collettivo di Hacker, Filosofi e imprenditori. Lavoriamo nel mercato AI e cybersecurity mentre ci facciamo domande sul mondo che verrà.

    Likely audience: ML engineers

  9. 9

    Substacklast post Jul 30, 2026

    On the A Line for the benefits and to grow up

    Likely audience: AI policy researchers

Latest posts from these newsletters

Browse all topics