Live data from Hacker News

Brave Browser has an AI assistant chat now

brave.com

1–10 of 25 posts

Re: Brave Browser has an AI assistant chat now

#4
> The model behind Leo is Llama 2, a source-available large language model released by Meta with a special focus on safety. We’ve made sure that user inputs are always submitted anonymously through a reverse-proxy to our inference infrastructure. In this way, Brave can offer an AI experience with unparalleled privacy.

> We’ve specifically tuned the model prompt to adhere to Brave’s core values

I wonder what does this tuned version represent?

Re: Brave Browser has an AI assistant chat now

#6

reverse-proxy for access to their server - they claim they have no records of individual IP that way

Nobody records your IP when you confess your sins to a priest either. The things you say in these contexts are self-doxxing. Never trust AI privacy guarantees!

Especially from Brave. I like and use their mobile browser, but at one point their useragent had my fucking phone model imprinted in it (these days it only betrays that my phone is running Android 9). Their idea of "privacy" is a joke.

Re: Brave Browser has an AI assistant chat now

#7

reverse-proxy for access to their server - they claim they have no records of individual IP that way

Nobody records your IP when you confess your sins to a priest either. The things you say in these contexts are self-doxxing. Never trust AI privacy guarantees! Especially from Brave . I like and use their mobile browser, but at one point their useragent had my fucking phone model imprinted in it (these days it only betrays that my phone is running Android 9). Their idea of "privacy" is a joke.

>at one point their useragent had my fucking phone model imprinted in it

That was pretty standard among browsers a couple years ago.

Re: Brave Browser has an AI assistant chat now

#10
post #4

> The model behind Leo is Llama 2, a source-available large language model released by Meta with a special focus on safety. We’ve made sure that user inputs are always submitted anonymously through a reverse-proxy to our inference infrastructure. In this way, Brave can offer an AI experience with unparalleled privacy. > We’ve specifically tuned the model prompt to adhere to Brave’s core values I wonder what does this…

Hopefully the prompt gets leaked soon enough. It's probably just generic "be helpful" with maybe a dash of "buy crypto, there are no other browsers".
Post reply on HN