AI Connections
How it works: 16 terms from this Drop's news make 4 groups of 4. You can make 4 mistakes. About 3 minutes. Your streak lives only on this device.
Mistakes left: 4
Show the groups
OpenAI's embedded safety assessors
Apollo · METR · Transluce · Goodfire
OpenAI says these four groups get deep access, such as checkpoints, source code and monitoring systems, for its first phase of embedded assessments.
From the Drop: OpenAI gives Apollo, METR, Transluce and Goodfire deep access for safety assessments
Named in the US chip rule's license list
Malaysia · Singapore · Indonesia · Thailand
TechCrunch reports the interim rule adds license requirements for chips sent to these countries, along with Brazil.
From the Drop: US rule requires licenses for remote access to AI chips by firms in China
Anthropic's disclosure clock
90 days · 30 days · 45 days · 15 days
Anthropic says severe bugs go from 90 to 30 days before disclosure, other bugs get 45, and it will escalate silent maintainers after 15 days.
From the Drop: Anthropic cuts its bug disclosure window from 90 to 30 days for severe flaws
Buying into Anthropic's IPO
AMD · Broadcom · Micron · TSMC
CNBC reports these chipmakers are among the companies planning to invest $26 billion at Anthropic's IPO price.
From the Drop: Altimeter aims to put $5 billion into Anthropic's IPO, Gerstner tells CNBC
Games are written by AI from this Drop's sourced items. Scenarios are illustrative.