Anthropic Shocking Privacy Updates

So, Anthropic plot twist. The AI that promised privacy now wants your diary. So, have you seen the latest updates to Anthropic? I guess all of their products. They now want to keep your chat data, chat history, your code history, and claw code as well for five years and train off your data too. So, it's not just like holding on to it, you know, for you know, I don't know what you couldn't call it like HR or whatever. um is literally they're training off your data. And this is re a stark contrast kind of to a company that in recent months has been releasing so many articles about you know just privacy in general and and you know making sure that AI is built for good and this and that and you've basically got until the 28th of September you should be presented with some sort of screen to to accept this. Now I I have because we've seen what Siri can can't do because of all the privacy stuff. I want a better product from Claude. I'm happy them seeing what I'm coding. Again, this is maybe a developer perspective with all this, but they're literally Yeah. They're going to be keeping your data for up to five years. So if you say yes to it now, that means they'll have your data, all of your data until the year 2030, which is just crazy. Funny to see your developer mindset as well. It's like just take all my life, take my data, my code. I I I mean, they already have it. I want a better product. I want a I'm le I'm I I see the light. I'm I'm I want a better product and I don't mind sharing my data to give better product. you know, Streamyard. We use Streamyard to do all this, that, and the other. Every single time we do a show, I click generate AI. We never use them, but I want that to get better because if we can AI generate a bunch of shorts for you guys and pump them out, um, then that's just better for everyone. Uh, so I just I just click it because I'm whether I don't know whether they're training on it or not, but I just want that to get better and better and better. So, what's your perspective? I mean, are you are you saying yes or do you use you use core, don't you? Um, very little, but I would say yes. Cursor. Yes, I do. I do. I wonder if I wonder how this does it with through cursor then. Do you think cursor you need to accept through cursor? I didn't notice anything new through cursor. I've been using cursor and especially like their claw models. But I would say yes to it. I mean, if they have my data for 30 days or 5 years, I don't think it's going to make a difference. They they have my data anyway. And even even if they don't store the 30-day version, you know, for longer, they could train, you know, on it and store it in a different version, right? Um, so I I don't care about this too much. It's it would be nice to not, you know, give away your data, but I guess this is the world we are living in and I don't feel like fighting against it to be honest. Yeah. At the end of the day, if we're getting better products from it, I mean a Google have always done this. You know, you they give away their products for free. Yeah. In return, we get good products. If it's not a good product, then whatever. I really I mean, don't get me wrong, I'm with you guys here when it comes to privacy. I really don't want this to bite me in the ass. I really, really don't. I want to have faith in humanity. This might be quite I haven't gone through this. This might be quite interesting to kind of go through all lang Why are you making this change? All large language models language models like Claude are trained using large amounts of data. Data from real world interactions. Now here's the thing. I actually don't think coding, you know, we've got all the data out there. If you ask an AI to build you an app, it's going to build it in React and it's going to build it using Tailwind and Shad CM. It's not it's not going to keep up with some of the latest trends, right? I think this will make sure they're ahead of the game because they're actually able to learn of what people are actually building or asking the AI to build. Granted, probably most people will just be happy with the React, you know, Express, you know, full stack application, but there will be developers who want to force inel or who want to um use their own systems using um Anthropic. Now, I wonder if there's anything to do with tool use because obviously when I use a tool, I just installed it the other day, but it's similar to context 7 and in that it gets documentation up to-date documentation to help write code because of course if I ask it to use a library that was released last week, then it's not going to know anything about that because I don't know when the cut off date is with Claude. I think it's September last year. Does the result from the tool come into play? So I've just searched tool and it doesn't show up anywhere but anyway let's um let's continue with this. Um when developer debugs code collaborating with AR model interaction offers valuable signals. So basically yeah I mean as you as you'd expect and why are you extending a development cycle span years basically 18 to 24 months ago they're they're you know they're up until so you won't get anything younger than 18 months ago or older. Yeah, keeping the data consistent through the training process helps make the models more consistent to models trained on similar data will respond reason and produce outputs in similar ways. I mean yeah there's nothing really here that's you know updated this to be honest but there is a there is a bit of backlash people are a bit worried about it particularly when it comes to data security you know again we reported last week that um Grock was was announcing that a bunch of things were being published to publicly now whether again I don't know I speculated whether that was actually a developer mess up or whether there was a communication issue and actually they've done nothing wrong but it was just not communicated that that would be indexable. So don't just share your bomb making history whatever the people were sharing. Um Otter AI as well a company that messed up on they actually got done a whoopsie and that they were actually storing people's data like there's just a there's just this worry that are they capable of actually storing data and without any compromisation without any breaches or anything like that. So there is a worry there. Um how is it with open AI and other companies like how long are they storing our data? Do you I think they're 30 days. I do think 30 days. But even if it's one day, can't they theoretically generalize? I don't know the process. I know what you mean. So once that data is in, they take it and then they're out, right? Is that what you're saying? Yeah. If Claude can get away with it, Mark Zuckerberg is like he's very happy because Claude is getting away with it and he will be asking for more now. Quite interesting. Let us know in the comments. What are you worried about AI? What do you what do you think AI is going to do with your data? What are you worried about? I'd love to know. We would love or like what is the private type of thing that you are talking to AI to about that you don't want to be published. Yeah. What are the questionable things that you are asking? This is part of a larger conversation on my show, Command AI, which we stream live every single week. We discuss the news and all things related to AI in the world of design and web. Catch us next week and join in the banter.