Can governments quickly and cheaply slow AI training?
I originally wrote this as a private doc for people working in the field -…
Recent Articles
View AllOpenAI delays ChatGPT’s ‘adult mode’ again
The feature, which will give verified adult users access to erotica and other adult content,…
Did I Catch Claude Cheating?
OverviewIn my API interactions with the Anthropic API I am finding what appears to be…
D&D.Sci Release Day: Topple the Tower!
This is an entry in the 'Dungeons & Data Science' series, a set of puzzles…
CHAI 2026 Workshop: Open Call for Posters!
To mark the Center for Human-Compatible AI's tenth annual workshop, we're casting a wide net…
AI Safety Needs Startups
Summary:Startups can become integrated in the AI supply chain, giving them good information about valuable…
More is different for intelligence
Why did software change the world?In the 1900s, much of the work being done by…
Your Causal Variables Are Irreducibly Subjective
Mechanistic interpretability needs its own shoe leather era. Reproducing the labeling process will matter more…
Mox is the largest AI Safety community space in San Francisco. We’re fundraising!
Summary: Mox is fundraising to maintain and grow AIS projects, build a compelling membership, and…
Self-Attribution Bias: When AI Monitors Go Easy on Themselves
PaperA common pattern in AI evaluation pipelines involves using an LLM to generate an action,…