1 min read
Add as a preferred source on Google

OpenAI Sets Four-Month Apollo Pilot to Probe AI Safety Controls

Six evaluators will examine whether OpenAI models try to undermine safeguards for their own or future models. The pilot’s start date has not been announced.

Six evaluators inspect an unmarked console in a quiet testing room / TokenPost.ai
Six evaluators inspect an unmarked console in a quiet testing room / TokenPost.ai

OpenAI will host six Apollo Research evaluators for a four-month review of whether its models try to undermine safety and control mechanisms for their own or future models. The pilot gives outside evaluators a defined period inside the company to examine model development and internal use.

Simon Yoon

Reporter

Simon Yoon reports on blockchain technology for TokenPost. Send corrections or tips to info@tokenpost.com.

Loading…