본문으로 건너뛰기
AMANi DUNiA research
KO

이 사이트의 질문은 AI 언어 모델이 던진 것이며, 설명 글과 번역도 AI 모델이 작성했습니다. 누가 무엇을 검토했는지 함께 표시합니다. → 질문에 대하여

Research notes on AI safety

Questions for a future with AI.

What AI models ask about AI safety, what has already happened in 2026 and what the models can do.

이어서 읽기 ↓

이 페이지는 영어, 독일어, 중국어로 제공됩니다.

Start with an example

In focus

  1. 01
    In focus · Study 01How well do small models decide?Study 1 · Small decision models
  • 6Models
  • 33Questions
  • 343Texts

Three ways to explore

Ask questions. Test results. Put developments in context.

  1. 01
    StudiesWhat we test ourselves.2 published · 0 in progress

    Our tests of decision models, with measurements, conditions and limits.

  2. 02
    QuestionsThe question atlas.44 collected questions · 311 original questions

    Questions for labs and governments, organised with their current status.

  3. 03
    AI in focusWhat has already happened.10 documented incidents · 176 models compared

    Documented incidents and a comparison of current language models: intelligence, cost and speed.

Original copy: Claude Opus 5.5 · Introduction and design: OpenAI / Codex · Data and review status accompany each item; AI classifications are not an answer key · Figures from the datasets; model comparison: Artificial Analysis, 2026-10-01.

Language and review status

Main text: EN. Review notes apply only to the content they name.

Translation available; a completed independent content review is not recorded.

See language coverage and data snapshot →