Hacker News Reader

Comment by zirgs | Hacker News Reader

One of the biggest drawbacks of ChatGPT is that OpenAI knows everything that its users are doing with it. Every prompt and its answer are being logged. Hackers might breach OpenAI systems and leak its data.

If you're Rockstar that's working on GTA 7 then you'll propbably want to keep all the AI written mission scripts, story ideas, concept art and other stuff like that on your own servers.

View on HN

... which is one of the strongest arguments for being able to run a large language model on your own hardware instead! https://til.simonwillison.net/llms/llama-7b-m2

Don't want to be vapid, but these are some cool guides! I know how to run these models but want to link my friends to guides to get started. Thanks!

> OpenAI retains API data for 30 days for abuse and misuse monitoring purposes.

They just changed this. It is now only 30 day retention - https://openai.com/policies/api-data-usage-policies

Data retention is kind of meaningless in this context since there's so many ways it is laundered/absorbed/analyzed while not technically violating whatever legalese they use this month.

Does that only apply to API usage and not ChatGPT the web app?

It would seem to, because the web app doesn't seem to expire your old chats.

My old chats all completely disappeared in the past few days . Quite disappointing actually. I had some valuable state in some of them.

Yeah, only the API. To opt out of the app, you’ll need to fill out a Google doc https://help.openai.com/en/articles/7039943-data-usage-for-c...

Isnt this the case for a lot of web products? Hackers can hack into Adobe and steal my prototypes. They can hack into my Dropbox and steal my files. They can hack into my Asana project and steal my roadmap

I agree with you, but I do think that people are overstating the problem. It's no worse than sticking your data on the cloud, and a huge portion of companies are doing that willingly already.

If I store my files on the cloud then I can encrypt them locally and upload them already encrypted. So that the cloud provider can't read them. OpenAI, on the other hand, can read everything that I submit to ChatGPT.

>Hackers might breach OpenAI systems and leak its data.

I think you should be more worried about OpenAI themselves instead of "Hackers".

There's also GDPR and similar legislation in the EU. OpenAI is based in the US and if your users are in Europe, you may end up sharing their data to a US entity. So you'd need to consider the legalities of that.

I have a growing list of use cases in the context of a SAAS app where I might want to use openai for various things. But this one could be a deal breaker with some of our customers.

Reply on news.ycombinator.com