Hacker News Reader

Comment by bertil | Hacker News Reader

To do that, you have to ask for users’ informed consent.

It makes no sense, UK still has the same consent laws as EU, and Bard is deployed there. Maybe an oversight? It does look like they don't know what they are doing.

It makes sense if Google believe the Eu appetite for fines is greater than the UK’s.

Could it be that UK, the country that spies and records what everyone does, isn't so big on privacy in practice and they thought the risk is less?

I wouldn’t exclude an oversight, given how they didn’t exclude Åland.

https://news.ycombinator.com/item?id=35914946

If you look further down that thread, people also highlighted they included more countries with “adequacy,” i.e., that the EU considers having a comparable privacy regime.

Asking for consent isn't enough. You also have to provide an alternative for users who don't want to consent.

The problem may be that you cant delete the data on request once it has been used to train a model.

A similar problem happens with all of their existing continuously trained models, like the ones that power their ad recommendation system.

They do comply with that. They can also comply with this.

I think it all boils down to opportunity costs. Implementing a system that will ensure that the models are fully retrained from raw corpus/logs within the required timeframe simply takes time. Time they don't have since openai and MS are pressing them in that front.

I think some teams at Google will work in the background to ensure that the whole shebang will be GDPR compliant while in parallel google signals to the markets that they are not a dying giant

Reply on news.ycombinator.com