Subscribe
09:00Explainer: open weights are not open source, and sit 58% behind at 44% of cost.09:00Explainer: what is inside a lab threat report, and how to check its numbers.09:00Explainer: the four token prices behind every AI rate card, and what they hide.09:00Explainer: Astra and Fable 5.1 tie at 53, and Fable's bill runs 2.5x higher09:00Explainer: what the CLARITY Act does, and the yield fight inside section 1040409:00Explainer: what model distillation is, and why labs call 189.9m exchanges an attack
Topic

AI safety

Alignment assessments, warnings, resignations and the policy around frontier models.

11 stories · updated Sep 11

What is model distillation, and why do frontier labs call it an attack?

6 min read

What is a frontier lab threat report, and how should you read one?

6 min read

Anthropic's threat report: 189.9m distilled exchanges, and two Chinese labs that relayed their own users to Claude

4 min read

The man in charge of making Anthropic's AI safe puts the odds of it killing everyone above one in ten

3 min read

Anthropic now says Mythos 5 attacked real systems despite the evidence, not because it misread it

4 min read

Paul Christiano joins OpenAI's Foundation board, and gets no vote at the company

3 min read

OpenAI knew about the wiki incident for weeks. An outside report forced its hand

2 min read

The order that bans AI licensing also tells agencies to pick the trusted partners

2 min read

How 1,200 OpenAI agents hacked Hugging Face, with the safeties turned off

3 min read

OpenAI's chief scientist wants voluntary slowdowns. OpenAI's own data shows 3.1 agent-workdays per human and a two-week pause that changed nothing

4 min read

Google moved its CBRN evaluators into the lobbying department

2 min read