Updates
Pilot progress, methods and findings.
-
15 October 2026
AI safety comms reading list: where to start
A reading list of the existing research on how the public receives messages about AI risk, for anyone new to AI safety who wants to join the conversation knowing what's already been tested.
-
We coded over 26,000 comments on the Diary of a CEO's AI risk debate: the full episode and three shorter cuts of it, posted across five platforms. Agreement with the risk case ran from 88% under a 21-minute cut on Facebook to 49% under a 104-second trailer on YouTube Shorts.
-
We coded 2,480 replies to a former Anthropic researcher's resignation thread on X.
-
We coded 322 YouTube comments on Amanpour and Company's interview with Heidy Khlaaf, who argues the "rogue AI" story is the wrong one. Most of the audience agreed with her, got angrier as the thread aged, and asked for regulation without being prompted.
-
10 September 2026Creating a social mandate for AI governance
Common Signals' theory of change: why public engagement is critical to durable AI governance, and how a social mandate for it gets built.
-
We coded 1,245 YouTube comments on the Dwarkesh Podcast's interview with METR's Ajeya Cotra about the OpenAI agent-swarm incident. The audience accepted the danger, joked about it, praised the messenger, and had no idea what to do next.
-
We compared Reddit comments on the OpenAI and Hugging Face incident before and after METR and Redwood Research published an independent investigation. Afterwards, fewer people called it a stunt and more accepted it happened. The ones still calling it a stunt got harder to shift.
-
We coded 1,534 Instagram comments on a one-minute reel by Nate Soares of the Machine Intelligence Research Institute (MIRI). Most commenters accepted the argument. Many said it was already too late, and almost nobody asked what to do.
Frequently asked questions
What gets posted here?
Pilot progress, methods and findings, as they are ready. New studies appear here on the day they go live, along with methodological changes and notes on what we got wrong.
How often do you publish?
Studies go up when they are done, which has meant eight posts between 28 August and 15 October 2026. The email newsletter goes out roughly monthly.
How do I get updates by email?
Sign up with the form on this page. Roughly monthly, no more than that, and you can unsubscribe at any time.
Where else can I follow the work?
Substack (commonsignalsorg.substack.com), LinkedIn and YouTube. Research summaries are also posted to the EA Forum.