Last week of September, another wave of high-voltage hysteria in the Internet regarding AI and the destruction of humanity emerged.
It all began with a viral post from ex-employee Jacob Coxon about the unstoppable AI development heading straight towards superintelligence. The thread ↗ suspiciously went viral for an account that appeared in January 2026 and had only a few posts. But after just a few reposts from popular handles, the post quickly got more than 170M impressions, getting endorsements from such famous people as Geoffrey Hinton ↗.
A couple of days later, instead of de-escalating things with a logical explanation, Dario Amodei decided to publish an essay titled “We Must Pace the Frontier” ↗, calling for self-regulation and slowing down the speed of frontier development. Right after posting that on X, direct competitors like Sam Altman and Elon Musk unexpectedly joined him calling for a coordinated pause.
The fear marketing cycle.
AI companies releasing new frontier models every three months need to stop being hypocritical. Appealing to “pace the frontier” while shipping a new model a couple of days later is incongruous at best. In a top of that using fear narratives as a marketing tool to sell API credits is way worse.
Unfortunately this we are in a geopolitical trap. The US government wants to maintain maximum leverage against advancements from China in dominating this technology stack. That leaves companies like Anthropic and OpenAI in a tight spot, caught between political pressure to accelerate and their own “ethics” to look responsible.
Can the AI actually destroy humanity?
As for me, current models are incredibly powerful and capable, but I feel, we are far away from the autonomous Skynet wiping out humanity.
Are current models capable of developing targeted cyberattacks on critical infrastructure? Yes. Are these models capable of designing some sort of specialized weaponry? Without a doubt. But the point is there is still a human behind the keyboard, not an autonomous and self-aware technology.
My take on this.
AI development is not going to slow down.
If there is any government intervention and safety regulations are enforced, we are going to see new releases much less often, and the model capabilities with every new version is going to decrease with time. But that does not mean the frontier development is going to stop, far from it. Companies are going to develop models at full speed behind closed doors, probably classified as restricted tech intended primarily for state actors and defense applications.
AI is essentially a tool, and it either builds or destroys depending on who uses it. Current threat to humanity is not some superintelligence awakening tomorrow; it’s the human typing a prompt.