these AI deployment companies would potentially have a lot of intimate knowledge about the real world and real people, either via the web interfaces or perhaps customer data from robots. The AI R&D orgs would not be allowed to have that kind of data
I am not intimate with the details of R&D of frontier models, but don't they use customer telemetry & data for making their models better? In this scenario, where would they be able to get data from their products?
Is them not being able to get feedback of their product to improve them is the intention?
It's a well known photography phenomenon that if you take many photos, you'll increase your chance of getting a really good one. I think that's why people are approaching less harshly to the lying part.