Collection

AI Safety and Alignment

How do we make sure powerful AI does what we intend? These sources range from a beginner-friendly newsletter and videos to the research forums and labs doing the technical work. Start at the top of the list and move down as the ideas become familiar. Some of these sources are technical research forums, so expect jargon further down the list. The safety newsletter and the videos are the easiest way in.

Chosen by our editors · updated 2026-11-21 · 7 sources
Open my wall
    • ASNewsletter

      AI Safety Newsletter

      Center for AI Safety's newsletter on the latest in AI safety.

      0.5/wk since Sep 1 · 11 days ago

    Pick 1

    AI Safety Newsletter

    Why we picked it: The Center for AI Safety's plain-language news round-up. The easiest way into AI safety news; expect a short, plain-language round-up of incidents, research and policy.

  1. Pick 2

    Robert Miles AI Safety

    Why we picked it: Videos that explain safety problems to everyone. Made for newcomers to AI safety; expect careful, often funny explanations of why aligning AI is hard.

    • MBlog

      METR

      Updates and write-ups on measuring frontier AI capabilities and risks.

      1.4/wk since Aug 26 · 1 day ago

      Steady

    Pick 3

    METR

    Why we picked it: Independent evaluations of what frontier models can do. Suits technical readers interested in evaluations; expect research reports measuring what frontier models can do.

    • RRBlog

      Redwood Research blog

      Research on catastrophic AI risks and ways to reduce them.

      1.9/wk since Sep 10 · 3 days ago

    Pick 4

    Redwood Research blog

    Why we picked it: Research on catching and controlling misbehaving AI. Best for readers already familiar with alignment; expect detailed research posts on AI control.

    • TCBlog

      Transformer Circuits Thread

      Anthropic's interpretability research on reverse-engineering language models.

      0.3/wk since Oct 29 · 23 days ago

    Pick 5

    Transformer Circuits Thread

    Why we picked it: Anthropic's research on looking inside models. For readers ready for technical depth; expect long interactive research articles on model internals.

  2. Pick 6

    Future of Life Institute Podcast

    Why we picked it: Conversations on AI risk and how to steer it. Suits listeners who care about long-term risks; expect long interviews with researchers and policymakers.

    • IAPodcast

      Into AI Safety

      Conversations exploring the field of AI safety.

      0.3/wk since Sep 29 · 12 days ago

      Steady

    Pick 7

    Into AI Safety

    Why we picked it: Interviews for people exploring the field. Good for people exploring a career in AI safety; expect friendly interviews with researchers and practitioners.

More Collections

Type a name, a topic or a format. Enter opens the full results.