Hacker News Reader

Comment by simonw | Hacker News Reader

I don't think that holds up.

Paying customers absolutely HATE the idea of their data being used to train AI models without their permission - this Zoom story is just the latest example of that.

Companies that try to do this will get burned. Zoom just got burned really badly, and I personally don't think they actually intended to even do this - they just didn't make it clear enough that they were NOT going to do it, which sparked a PR nightmare firestorm for them.

I think the incentives for companies are very much the other way round: if paying customers hate this, then the incentives are NOT to do it.

View on HN

I promise not to use any of the data I collect about you!

Except to improve services (ML training!), Advertising (We'll sell your data to advertisers!), and by government order (pick your favorite three letter agency.)

Alternatively: They'll just use your data anyway and not tell you about it. How is anyone going to prove their data was used as the source? It's like all the NDAs people sign when they join a company and they pinky promise not to use it at the next job where they land a big fat raise and promotion... suuure they aren't going to take what they've learned and improve upon it to try and get more promotions and raises in the future.

It can't be stopped.

> It can't be stopped.

Maybe. Maybe not. Hard to tell without trying.

I applaud the EU and California for giving it a go with their data protection laws. I really hope their crackdown on this stuff is effective.

> It's like all the NDAs people sign when they join a company and they pinky promise not to use it at the next job where they land a big fat raise and promotion

Uh no. An NDA would cover proprietary intellectual property, not tools everyone else also uses. Unless you're now working for the previous employer's competitor, it's unlikely that proprietary tech would ever be used. Working for competitors and partners is usually also forbidden for some period of time after leaving.

> How is anyone going to prove their data was used as the source?

1) whistleblowing 2) compliance audits (think soc2)

How would a soc2 audit uncover that?

They're asking for evidence that your admin accounts are reviewed for permissions needed each quarter, that you are doing your DR tests, that you are following documented change management processes, etc, not asking to look at what your running containers are doing.

There is absolutely no way that any compliance audit I can think of would or would even attempt to uncover that kind of info.

I hope you are right but I fear it will become a process of boiling the frog. Companies that are in the business of renting access to your data will get increasingly clever at moving in this direction in small incremental steps, providing some user benefit along the way.

We can agree that Zoom did a terrible job of rolling out their new terms, regardless of what their intention was. What other companies will learn from this is to improve the roll out.

Once local/private inference becomes more viable, there will be even more of an incentive for the companies who store unencrypted data to use it as a competitive advantage.

This. I wouldn't be surprised if people give up after a while of searching and migrating to alternatives to each service that started using or selling data for AI training.

ehhhh I think some customers would be ok with it if it meant discounts. If your software is good enough, you can tell customers to go elsewhere. The theory is that AI done well can be incredibly powerful ($$$) so you have to build it or risk being left behind.

Sure, consumers are likely to say yes to that - but I'm talking about companies here who are much more sensitive to what happens to their data.

> Paying customers absolutely HATE the idea of their data being used to train AI models without their permission

Paying customers will be the ones to resist for the longest maybe but that frog too will be boiled eventually.

Reply on news.ycombinator.com