One private company decides who gets the most powerful model

The AI moment has come that I have been the most afraid of.

Anthropic announced it will not release their Mythos Preview model to the public. It supposedly is too powerful.

It found 27-year-old security holes in an operating system that runs firewalls and critical infrastructure.

It found attack vectors in the Linux Kernel and in FFmpeg - the video encoder embedded in nearly every piece of software you use.

Vulnerabilities that survived decades of human review and millions of automated tests.

Think about that for a moment.

When you have an AI model capable of intruding the operating systems our entire infrastructure relies on - what do you do?

Give it to everyone? In the hands of bad actors, this would be a global disaster.

That's why Anthropic gave it only to a selected group of "good guys":Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks.

But there are thousands of organisations worldwide whose entire reason to exist depends on IT security - and they didn't make the list.

Here's the deeper problem: OpenSource will eventually balance things out technically. Similar powerful models are already being built.

But running a model at this capability level costs serious compute.

So even if an equally powerful open model exists, only capital-rich organisations will be able to afford to use it at scale.

AI has stopped being just a helpful tool. It is now capable of shifting power dynamics between countries, industries, and individuals.

And right now, one private company is deciding who gets to wield it.

I find myself more and more drawn to the idea of an off-grid farm with chicken, a few goats, 2 donkeys and permaculture.

Turns out that's not a joke - you can't hack a donkey.

But for those of you still running on servers:

The most interesting thing about this list of "good guys" is that it was never yours to question.