Apollo Research to test AI safety from inside OpenAI

Six Apollo Research staff will work inside OpenAI for four months to assess whether models try to undermine oversight. The pilot gives evaluators limited access during training and internal deployment, and Apollo plans to publish two reports.

Sources 1
Get the app vidgetc Tech, gaming & AI news — always at hand Google Play · Soon App Store
Discuss

Comments

Next articleAMD completes its acquisition of World Labs→
Start typing to search