No, AI Won’t Kill Us
No, AI won’t kill us.
But fear of AI can still be extremely useful to the companies building it.
In September 2026, Anthropic CEO Dario Amodei called for slowing the development of frontier AI. Sam Altman agreed. Elon Musk agreed too.
Suddenly, the people leading the AI race wanted everyone to slow down.
AI could kill us, but it is still available for sale?
The media cycle began with a dramatic warning from former Anthropic researcher Jacob Coxon. His message that AI could “kill us all” received more than 100 million views.
It was presented as the spontaneous warning of a researcher who had just resigned.
It was not spontaneous.
Coxon spoke with The Wall Street Journal before publishing his message. He organized a group of people to amplify it. The post was immediately supported by Anthropic employees, followed by interviews with Wired, Axios, CNN, NBC and Fox News.
This was a public relations rollout.
There is no public evidence that Anthropic organized it. But it was clearly prepared and designed to receive maximum attention.
Four days later, Dario published We Must Pace the Frontier, calling for industry coordination, government intervention and limits on the speed of AI development.
That raises an obvious question: why did Coxon resign when he fundamentally agrees with Anthropic’s CEO? If both want the same coordinated slowdown, what exactly is the disagreement?
Sam Altman and Elon Musk quickly supported Dario.
Then, a few days later, Dario and Sam appeared at Salesforce’s Dreamforce, where the focus was once again on selling AI products, partnerships and the future of enterprise AI.
So which is it?
Is AI so dangerous that it could kill all humans?
Or is it safe enough to sell to every company in the world?
If your model is unsafe, don’t release it
The solution is actually simple.
Nvidia CEO Jensen Huang made the same point at Dreamforce 2026. “If you build a product or a service and you’re not confident in its functionality, capability, or safety, then don’t release it,” he said. Safety is the responsibility of the company building the product, not an excuse to slow down everyone else. Source: ITPro
If Anthropic believes one of its models is unsafe, Anthropic should not release it.
If OpenAI believes one of its models is unsafe, OpenAI should not release it.
If xAI believes one of its models is unsafe, xAI should not release it.
Why should every other AI lab, startup, researcher and open-source developer become responsible for the fact that these companies continue releasing models they claim they cannot fully control?
They built the models.
They control the infrastructure.
They decide when to release them.
They sell access to them.
The responsibility belongs to them.
They are leading the race
Imagine a sprint.
OpenAI and Anthropic are far ahead. Then Kimi, GLM, DeepSeek, Qwen and other open-weight models begin catching up, delivering similar performance for a fraction of the cost.
Suddenly, the leaders announce that the race is dangerous. They want new rules, mandatory testing and government intervention before everyone else can continue running.
That is a convenient way to protect their lead.
Open models threaten the business model of closed AI labs. Anyone can download them, run them locally or offer them through a cheaper service. As the performance gap closes, intelligence becomes less scarce, prices fall, margins shrink and customers become less dependent on OpenAI and Anthropic.
This also threatens their valuations. Anthropic is preparing for a possible 2026 IPO, while OpenAI has postponed its listing until 2027. Their valuations depend on investors believing they will control a large part of the future AI economy.
Anthropic says it does not want to ban open-weight models (lol). But mandatory approvals and release requirements could make powerful open models much harder to publish.
The companies currently leading the race would get to design its rules while protecting their scarcity, margins and valuations.
As open models and personal hardware improve, the cost of intelligence moves closer to zero.
That is good for everyone, except companies whose business depends on renting it.
Two possible futures
Imagine the first future.
You own your AI.
It runs on your Mac, phone or another personal device. Your files stay with you. Your conversations remain private. The model works offline and cannot be removed from your device remotely.
You decide which model to use.
You decide what it can access.
You decide when to replace it.
Intelligence becomes accessible to everyone. Privacy becomes accessible to everyone. AI becomes something you own instead of something you rent.
Now imagine the second future.
The most capable AI belongs to OpenAI, Anthropic and a few other companies.
You pay every month to access it. The provider decides which model you receive, how much it costs and what it is allowed to answer.
The best intelligence is reserved for the richest companies and selected partners.
Everyone else receives a smaller, slower or more restricted model.
The provider decides what is safe.
The provider decides who is trusted.
The provider decides what you are allowed to know.
A company would be able to decide that one user is allowed to access scientific knowledge while another is not. Legitimate researchers could receive restricted or degraded answers because a private provider has decided they are not trusted.
That is not intelligence for everyone.
It is intelligence for the richest and censorship for everyone else.
Which world do you prefer?
We know which one we want.
We want a world where people own their AI.
That is why Fenn uses open-source models that run locally on your Mac by default.
Fenn helps you find anything you have saved, seen or heard on your Mac. Your personal data stays on your device. Your files are not sent to OpenAI, Anthropic or another cloud provider.
AI should not become another resource controlled by a handful of companies.
It should be accessible to everyone.
If the leading AI labs believe their models are unsafe, they should stop releasing them.
They should not use their own lack of control as a reason to control everyone else.