94 comments

[ 3.5 ms ] story [ 79.1 ms ] thread
Just opened Claude app on Mac and saw a popup asking me if it's ok to train on my chats. It's on by default. Unchecked it.

I think Claude saw that OpenAI was reaping too much benefit from this so they decided to do it too.

You can opt out, but the fact that it's opt-in by default and made to look like a simple T/C update prompt leaves a sour taste in my mouth. The five year retention period seems... excessive. I wonder if they've buried anything else objectionable in the new terms.

It was the kick in the pants I needed to cancel my subscription.

I cancelled my subscription as well because of the opt-in by default.
(comment deleted)
Feeling cute, might put all your private conversation into my public LLM later
In my opinion, training models on user data without their real consent (real consent = e.g. the user must sign a contract or so, so he's definitely aware), should be considered a serious criminal offense.
TBH I’m surprised it’s taken them this long to change their mind on this, because I find it incredibly frustrating to know that current gen agentic coding systems are incapable of actually learning anything from their interactions with me - especially when they make the same stupid mistakes over and over.
What a surprise, a big corp collected large amount of personal data under some promises, and now reveals actually they will exploit it in completely unrelated manner.
(comment deleted)
Ugh Anthropic please don't become the villain.
I think there is amazing signal inside the chat logs. Every idea or decision taken can be analyzed in hindsight 20 messages later, or days later. Eventually a feedback signal or outcome lands back in the chat logs. That is real world idea validation. Considering the hundreds of millions of users and their diverse tasks that collect across time - this is probably the most efficient way to improve AI. I coined it the human-AI experience flywheel.

To make it respect user privacy I would use this data for training preference models, and those preference models used to finetune the base model. So the base model never sees particular user data, instead it learns to spot good and bad approaches from feedback experience. It might be also an answer to "who would write new things online if AI can just replicate it?" - the experience of human-AI work can be recycled directly through the AI model. Maybe it will speed up progress, amplifying both exploration of problems and exploitation of good ideas.

Considering OpenAI has 700M users, and worldwide there are probably over 1B users, they generate probably over 1 trillion tokens per day. Those collect in 2 places - in chat logs, for new models, and in human brains. We ingest a trillion AI tokens a day, changing how we think and work.

All models are trained on data obtained without consent. Now the “good guys of AI” want users to add their chats to the stack and the users balk? Hilarious.
never provide these services with personal data / identifiers.
It's not just chats, is it? It says "coding sessions" too.
Whatever your secret projects and plans are, they’re now available to train up your competitors on what you’re doing!
GO to setting and under privacy you can change the default. (they make it pretty hard to find). Also note "To help us improve our AI models and safety protections, we’re extending data retention to 5 years."
> unless users actively opt out by September 28th

Looks like there is an opt out option. Curious about the EU users - would that be off by default (so: opt in)?

Not a surprise. All the major players have reached the limits of training on existing data—they’re already training on essentially the whole internet plus a bunch of content they allegedly stole (hence various lawsuits). There haven’t been any major breakthroughs in model architecture from the major players recently and thus they’re now in a battle for more data to train on. They need data, and they want YOUR data, now, and are gonna do increasingly shady things to get it.
If you don't like this, use local AI.
How long before some poor kids chocolate pudding fetish makes it to the model and they get bullied at school for it? Treat LLM chats as public.
Frankly I don't mind much if they use my js debugging conversation to train the next models; but there should be a way to mark specific conversations as private (possibly even an "incognito mode") to exclude them from any training or external access.
As soon as Apple said it woll ship it in Xcode? what a coincidence!
What about paid corporate accounts?

If I'm not paying for something, I presume this is the kind of thing that's happening, so this isn't newsworthy to me. Is it also applicable for paid and paid corporate accounts?