Are GPT-5 models more "ideologically neutral"?

Are GPT-5 models more "ideologically neutral"?
(New Location!)

The U.S. AI Action Plan declared a national push for leadership in artificial intelligence, mandating federal agencies to use only “ideologically neutral” models. Soon after, OpenAI released GPT-5 and secured a government contract, prompting questions about whether these models have shifted ideologically.

Matt is a scientific Python programmer at Equinor, and in an attempt to see if there is evidence for the kind of ideological shift the US administration is asking for, he has reproduced the work of "Cultural bias and cultural alignment of large language models" (doi.org/10.1093/pnasnexus/pgae346) paper, and extended it to more models.

(Spots are limited and people end up on a waitlist, so register early!)

We meet at Visere office (NEW LOCATION) in Chr. Michelsens Gate 1. If you are struggling to find it, send me a message here on the meetup (Dalibor Jovic) with your phone number, and I will help you up.