Bill asked me to compile a list of sources that one could refer to when keeping up with this rapidly expanding field. I’ve written a long blog post about Keeping Up (one of my favourite activities), but that’s more about what it is like to keep up than any helpful lists. So here we go…

First of all, we have to consider what we are keeping up with. In this case it is AI Safety. AI Safety wasn’t a well-defined field (probably still isn’t, really) but it has become much more coherent in the last little while. We won’t just confine ourselves to actual AI Safety, however, since you still need to have the context of “what is happening in AI/frontier labs” and you need to consider the AI Industry and financial implications, you need to consider governance and regulation and you should also be aware of those who are skeptical of the whole enterprise. It’s not really concentric circles, more like Venn diagrams or something.

The other issue to keep in mind is that many of the sources that one might follow are themselves aggregators. So you are going to see the same thing multiple times. From my point of view that’s just the cost of doing business, and its also a useful bell-weather for seriousness. In other words, if you read the same story everywhere (e.g., The Incidents™) then you might know that it is something big.

I already do a lot of this, and have my own lists, although I have never shared them withanyone else, nor provided commentary. Back in the early days of the web and the first blogs, it was common for people to create a “blog roll” that sat alongside their blog posts. It provided links to what they were reading and - to be honest - it was a simple tool for yourself as a blogger, since you spent a lot of time looking at your own page, tinkering with it and reloading it, so why not have your favourites right there on the same page. It also was a kind of early system of “likes” in that you were showing who you thought worthy of sitting alongside your content.

Maybe the first place to start is at the edges. There are some people who (still) have not taken the red pill (see my post about AI pills) and we can’t forget about them, since they still shape discussion in some ways.

Skeptics

Cory Doctorow (also a columnist for The Guardian): https://craphound.com and https://pluralistic.net

Ed Zitron https://www.wheresyoured.at (has premium option)

Melanie Mitchell https://melaniemitchell.me (scholarly approach)

Gary Marcus https://garymarcus.substack.com

Institutions

Redwood https://blog.redwoodresearch.org

FLI - https://futureoflife.org

Apollo Research https://www.apolloresearch.ai/blog

UK AISI https://www.aisi.gov.uk

Canada CAISI https://ised-isde.canada.ca/site/ised/en/canadian-artificial-intelligence-safety-institute

US CAISI https://www.nist.gov/caisi see also https://www.nist.gov/artificial-intelligence

Commentators

Zvi (brings together many of the Twitter/X AI Safety community, so you don’t have to follow them individually)

Newspapers/Magazines

• Guardian (has an AI category) https://www.theguardian.com/technology/artificialintelligenceai
• NY Times (requires a subscription) https://www.nytimes.com/section/technology
• Atlantic (covers AI well): https://www.theatlantic.com/search/?q=artificial+intelligence

Blogs/SubStacks

• Zvi https://thezvi.substack.com
• AI StopWatch https://aistop.watch
• Humans on AI: https://p3humansonai.substack.com
• Forethought https://newsletter.forethought.org
• LessWrong https://www.lesswrong.com
• ArXiv https://arxiv.org/list/cs.AI/recent especially cs.MA (multi agent systems) https://arxiv.org/list/cs.MA/recent
• The Next Web https://thenextweb.com
• Techdirt https://www.techdirt.com/search/?q=AI
• MIT Technology Review (subscription) https://www.technologyreview.com/topic/artificial-intelligence/
• Nature.com (machine learning category) https://www.nature.com/subjects/machine-learning
• Atlantic Council on AI https://www.atlanticcouncil.org/issue/artificial-intelligence/
• RAND on AI https://www.rand.org/topics/artificial-intelligence.html
• The Conversation (various academics) https://theconversation.com/ca/search?q=artificial+intelligence+

Thoughtleaders/scientists/CEOs

• Bengio - https://www.facebook.com/yoshua.bengio
• Hinton
• Le Cunn
• Hassibis
• Amodei
• Zuckerberg (!)
• Anthony Aguirre - https://x.com/AnthonyNAguirre https://www.anthony-aguirre.com

Gemini’s recommendations

Here is a curated list of top-tier AI safety resources, broken down by category so Bill can pick the depth and perspective that fits him best:

1. Dedicated AI Safety Newsletters & Digests

• The AI Safety Newsletter (by CAIS): Published weekly by the Center for AI Safety, this is arguably the best single roundup covering technical developments, public policy, and industry news with concise summaries.
• Import AI (by Jack Clark): Written by Anthropic co-founder Jack Clark, this weekly newsletter covers frontier machine learning research, safety implications, and global governance with deep technical and policy analysis.
• ML Safety Newsletter (by Dan Hendrycks): Focuses specifically on technical research papers in machine learning safety, robust alignment, monitoring, and systemic safety.
• AI Alignment Newsletter Archive: Curated historically by Rohin Shah (Google DeepMind), this is a foundational archive covering core technical alignment literature and arguments.

2. Community Hubs & Discussion Forums

• AI Alignment Forum: The primary hub where technical researchers, theoretical alignment thinkers, and governance experts publish research write-ups, debates, and literature reviews.
• LessWrong (AI Section): A major forum where many AI safety concepts originated; heavily overlaps with the Alignment Forum and features broader discussions on forecasting, rationality, and existential risk.
• AISafety.world: An interactive map of the AI safety ecosystem, tracking active organizations, research labs, funders, and community initiatives.

3. Governance, Policy & Incident Trackers

• Center for Security and Emerging Technology (CSET): Provides data-driven analysis and reports on AI policy, compute tracking, export controls, and national security implications.
• AI Incident Database: A dedicated collection tracking real-world failures, harms, and misalignments of AI systems in production.
• OECD AI Policy Observatory (OECD.AI): Great for tracking international regulations, country-level strategies, and standardized safety metrics.

4. Frontier Labs’ Safety Blogs

To track policies and safety releases directly from the frontier builders:

• Anthropic Research & Policy: Covers mechanistic interpretability, responsible scaling policies (RSP), and constitutional AI.
• OpenAI Safety & Alignment: Features preparedness frameworks, evaluations, and red-teaming reports.
• Google DeepMind Responsibility & Safety: Covers safe deployment, socio-technical risks, and capability evaluations.

Gemini’s lisk of think tanks

Here is a curated overview of prominent think tanks, research institutes, and policy centers dedicated to AI safety, governance, and frontier risk, categorized by their primary focus.

1. Major Policy & Security Think Tanks

These institutions mirror the RAND model, focusing heavily on geopolitics, national security, regulatory strategy, and supply-chain governance.

• Center for Security and Emerging Technology (CSET) (Georgetown University)
  • Focus: Data-driven policy research on AI hardware, compute tracking, national security implications, and international talent flows (particularly US–China dynamics).
• Center for a New American Security (CNAS)
  • Focus: Technology and National Security Program exploring military applications of AI, strategic stability, autonomous weapons, and export controls.
• Center for Strategic and International Studies (CSIS)
  • Focus: Houses the Wadhwani Center for AI and Advanced Technologies, concentrating on global tech competition, standards-setting, and public policy frameworks.
• Brookings Institution
  • Focus: The Artificial Intelligence and Emerging Technology (AIET) initiative covers domestic regulatory frameworks, algorithmic fairness, economic disruption, and international governance.

2. Dedicated AI Governance & Catastrophic Risk Institutes

These organizations specialize specifically in frontier model safety, existential/catastrophic risk, and long-term regulatory mechanisms.

• Centre for the Governance of AI (GovAI) (Yale)
  • Focus: High-level strategic research on international governance treaties, frontier model compute thresholds, lab self-governance, and long-term trajectory analysis.
• Center for AI Safety (CAIS) (San Francisco)
  • Focus: Technical safety benchmarks, systemic and catastrophic risk assessments, compute governance, and policy advocacy.
• Future of Life Institute (FLI)
  • Focus: Broad existential risk policy, international AI treaty design, autonomous weapons mitigation, and direct policy advocacy (notably behind key open letters on AI risk).
• Centre for the Study of Existential Risk (CSER) (University of Cambridge)
  • Focus: Interdisciplinary analysis of extreme risks, including catastrophic AI, biosecurity intersections, and resilient governance structures.
• Epoch AI
  • Focus: Quantitative research and forecasting on compute trends, algorithmic progress, training hardware scaling, and timeline modeling.

3. Evaluation, Technical Standards & Societal Impact

Organizations bridging technical audits, red-teaming benchmarks, and broader societal impacts.

• Ada Lovelace Institute (UK/EU)
  • Focus: Independent research and deliberative body focused on regulatory mechanisms, public interest technology, EU AI Act implementation, and algorithmic accountability.
• Model Evaluation and Threat Research (METR) (formerly ARC Evals)
  • Focus: Designing empirical evaluations and red-teaming protocols to test autonomous capabilities, cyber-offense, and self-replication risks in frontier models.
• AI Now Institute (New York)
  • Focus: Critical analysis of corporate concentration, compute monopolies, platform accountability, and regulatory enforcement.

Quick Comparison