We lean on AI more. We trust the companies behind it less.

Half of American adults now use an AI chatbot. In 2024 it was a third. One in four use one every day. (Pew)

Now the other number. Only 27% trust businesses to use AI responsibly, down from 31% last year. Among people 18 to 29 it's 20%, down from 30%. (Gallup) And 59% doubt American companies will build it responsibly. (Pew)

We depend on something we don't trust the owners of. And that gap is about to get a new pressure on it.

Anthropic and OpenAI have both filed confidentially to go public. We don't have to guess what happens next, because there's a name for it. Cory Doctorow calls it enshittification: first a product is good to you, then it's bent toward whoever pays the company, then it's squeezed for shareholders.

In AI it looks specific:

You rent a name, not a model. In February, OpenAI retired GPT-4o. About 800,000 people used it. They didn't get a vote.
Free tiers get lighter models and tighter limits.
Ads started appearing in ChatGPT in February.
Terms about training and data retention have changed after people signed up.

Nobody had to be a villain for this. Running these models costs hundreds of billions, only about 5% of users pay, and soon the owners answer to public markets. That's the whole mechanism.

The fix isn't a better company. It's owning the thing.

Open models now run on a Mac. That's what Stratumus is for: AI you own, on the Macs you already have, simple enough that you don't need to know what a model is. You pick the model. It doesn't change unless you change it. Nobody sells ad space next to your answers.

We're in testing and early access is open. Come see what it's like when the AI is yours.

Sources & further reading

← Back to the Journal