The Register · Thomas Claburn ·
OpenAI says new monitoring and security safeguards will add 20% compute overhead to monitored inference workloads; costs will not be passed on to customers
Expanded multistage chain of thought monitoring makes frontier model work more expensive
In this story OpenAI
Lead Source
Coverage
More
Discussion
@sama @andrewcurran_ @alexeheath @thezvi @hlntnr @openai @eliebakouch @gdb @daniel_mac8 @jun_song @mattshumer_ @scaling01 @bahradx @gabriel1 @joshua_saxe @repligate @tenobrus @yonashav @quantum_geoff @ross__hendricks @emollick @rand_longevity @emostaque @bubbleboi @chrisgpt @edzitron @jason @wholemars @kimmonismus @matthewberman @jasoncrawford @w01fe @miles_brundage @ziwenxu_ @jachiam0 @peterbarnett_ @deredleritt3r @thestalwart @mariushobbhahn @justalexoki @ns123abc @kelseytuoc @edzitron @merettm @deredleritt3r @sama @andrewcurran_ @bindureddy @iruletheworldmo @rohanpaul_ai @pigeon__s @lu_sichu @connortalksai @_nathancalvin @suchenzang @ananth7e @hesamation @j_asminewang @timjayas @mark_k @frenbilt @laz4rz @spoonedher @notjazii @xyhan_ @sicedricxd @rivonn @_nathancalvin @tokengremlin @rohit3a @dieaud91 @gabgarrett @kimmonismus @edzitron