[stouset]

This argument would be a lot more compelling if there was literally any company out there with an interest in doing this sort of thing that had even remotely comparable a track record to Apple.

OpenAI, Anthropic, Google, Meta, Perplexity, and everyone else in this space are openly brazen about how much of your information they want to consume—however possible—to store in perpetuity and use for whatever future purpose they want.

[Rohansi]

In this space, sure. AI only exists and is improving because of extensive data collection. That's why Apple licensed models from Google. Anyone can go and download and run open-weight models, and many allow commercial use. If Apple opened it up I'm sure you would see new options which respect user privacy pop up.

Just look at Apple's stance on third-party web browsers to see what I'm talking about. There are browsers which respect user privacy, have good security, etc. but Apple uses the same excuse there: only Apple can be trusted to do it right.

[stouset]

Now the issue becomes opening the door only to the few entities that can be trusted. It is much easier and cheaper, and carries considerably less risk to simply not.

This is one of those situations where multiple overlapping incentives all point the same way. In this case, Apple’s incentives align with those of user privacy and security, and so I am willing to give them a lot of leeway. Particularly since they seem to be the only large player taking this approach.

[Rohansi]

> It is much easier and cheaper, and carries considerably less risk to simply not.

Which benefits Apple because it gets rid of competition on their platforms.

> Apple’s incentives align with those of user privacy and security, and so I am willing to give them a lot of leeway

At the cost of your freedom. Sure, it's not that bad now, but I'd be worried about the future of macOS because it's the least restrictive Apple platform right now.