Skip to content
About AISafetyHot (AI Safety HOT)

New vulnerabilities, risks, attacks, defenses and evaluations emerge every day.
Keep up with them here.

AISafetyHot (AI Safety HOT) collects AI safety news and research from 1103 sources, updated daily.

AI Safety HOTHow it works

How information comes togetherRecent updates
OECD.AI（政策与事件监测）
GPAI expert group: investors hold the key to the future of trustworthy AI—are they ready?
Index · Today · Oct 12
论文追踪
SWE-Critix: A Post-Trained General-Purpose Verifier for Evaluating Coding Agent Trajectories
Post · Oct 9
论文追踪
EviRepair: Evidence-Driven Agents for Reliable Vulnerability Repair
Post · Oct 9
AI Safety HOTAll updates
OECD.AI（政策与事件监测） · Indexed
GPAI expert group: investors hold the key to the future of trustworthy AI—are they ready?
Index · Today · Oct 12

News sources, titles and dates move through the flow into the website. The lines illustrate collection, indexing, selection and publication.

Browse all updates and live monitorRead today's selected newsBrowse all papersRead the latestAI Safety Daily
  1. 01

    Collect

    1,103 sources

    Follow laboratories, researchers, academic communities and policy institutions.

    X 76 · RSS 173 · Web 26 · API 3 · Browser collection 825

  2. 02

    Index

    10.4K updates

    Index news and papers by category and bring reports on the same event together.

    1,444 items indexed in the past 24 hours

  3. 03

    Selected

    1,384 selected items

    Focus on substantive research, evidence and important developments.

    212 items selected in the past 24 hours

  4. 04

    Publish

    19 daily issues

    Daily, weekly and monthly reports review noteworthy content.

    Also available through RSS, API and MCP

20.2K visitors

Thank you to everyone who has visited

About the author

Oscar (Yutao) Wu

Hello, I am a PhD student studying AI safety. Seeing clickbait and exaggerated news every day often makes me anxious and disrupts my research. I wanted to collect the field’s news and papers, present them in plain summaries, and keep them up to date. I think agents are well suited to explaining what a news story says and what a paper studies. Checking this site each day now feels much easier.

We also plan a new series, “The story behind each paper.” From traditional journals and conferences to arXiv, and from CMT to OpenReview, the ways we share and discuss research have progressed considerably. I would like to take another step by inviting AI safety researchers to talk about why they studied a problem and their experiences working on the paper. I hope these stories will give new students useful perspective and help worthwhile work reach more readers. If you would like to share your story, please contact me by email.

Finally, thank you to AIHOT, Security of AI, and the WeChat publications, open-source frameworks and communities that provided inspiration. :)

Support this project

The site is free and requires no account. Servers and model calls have monthly costs. If it is useful to you, you are welcome to buy the author a coffee.

微信 QR code
WeChat
Scan to buy the author a coffee

Readers in China can scan with WeChat and choose any amount.

Buy Me a Coffee
Contributors

AI Safety HOT is a summary aggregator and reading index. Copyright in original content belongs to its respective sources. Source owners can request corrections, removal or display changes on the Message boardand choose “Correction / removal”.