This analysis was first published in SvD Näringsliv, in Swedish, on June 5th, 2026. This piece was translated from Swedish by Claude. Some phrasing may differ from a human translation.
Anthropic is calling for a global pause in AI development – and pressing on the industry’s most sensitive nerve. But there may be other reasons than concern for the world behind these warnings.
There is no finer way to express yourself in Silicon Valley than to write an essay.
Some have even turned it into a marketing strategy – as AI company Anthropic has.
By writing variations on doomsday essays, they have warned about AI development for many years. And raised hundreds of billions of kronor in venture capital along the way.
Reactions were strong after Anthropic on Thursday called on the outside world, and its own industry in particular, to pause AI development. “It would be good for the world to have the option to slow down or temporarily pause frontier AI development,” two employees noted in a blog post from the company.
The message presses on a sensitive nerve. Is AI becoming a kind of Terminator? Is humanity’s future threatened?
Narratives like these have defined AI development for many years. But it’s worth looking a little more closely at where they come from – and what possible motives may lie behind them, beyond concern for our shared future.
Let’s look at a timeline of what Anthropic has said previously on exactly these questions – and how they have since acted as a company.
March 2023: A blog post addresses three possible scenarios for the future. In the pessimistic version, Anthropic intends to “raise the alarm” – hoping that institutions around the world would then try to stop AI development. They also write that it may be difficult to notice when the pessimistic scenario kicks in, and that the world should assume that it has until proven otherwise.
May 2023: Anthropic raises 450 million dollars in venture capital from, among others, Google.
September 2023: Anthropic publishes a policy on “responsible scaling” – a framework for “managing catastrophic risks from advanced AI systems.”
March 2024: Amazon invests up to 4 billion dollars in Anthropic.
October 2024: Anthropic publishes guidelines for how governments can reduce “catastrophic risks.”
November 2024: Amazon increases its investment in the AI company to 8 billion dollars.
And finally – the big essay, by Anthropic’s CEO and founder Dario Amodei in January 2026. There he lists, among other things, that AI is the biggest threat to national security in a century – perhaps ever. That jobs will disappear. That AI models exhibit signs of extortion and cheating. And so on.
This could go on, but the point is hopefully already clear. These warnings are regular – yet shortly afterward, their own AI development accelerates with fresh billions in the bank. Despite, apparently, the world being on the verge of collapse.
Accelerate and brake, in turns. It looks a little strange.
In The New Republic in May, a similar tendency was noted. There you can read that it would be “easy to dismiss this as hypocrisy, but the situation is more structurally tragic than that.”
The reasoning was that on the unregulated market that exists, it wouldn’t be possible to stop development even if that was genuinely what one wanted. How would you ensure the stop? What would all the investors say? Would China stop developing AI just because the US chose to do so?
The questions are many. And despite incessant communication about how serious the risks of AI are, they haven’t received any clear – or at least realistic – answer. But positioning their company as a responsible social actor has been an extraordinarily successful strategy for Anthropic.
When the outside world reacts with alarm to these statements, it is worth remembering this context. The underlying anxiety in society seems to search for evidence of its own thesis – and what better than to hear it directly from those driving the development? They are, after all, the experts in the field.
Continuously raising the alarm about risks doesn’t necessarily mean those risks aren’t real. Nor that their concern is being used cynically. But it is a reminder that we seem to have outsourced these questions to a single party – the very largest tech companies in the world – without being able to form a qualified judgment of our own without them. That is an enormous problem we face here and now.
From the outside, it looks as though Anthropic is crying wolf. And like the boy in the famous fable, this creates a situation that ends badly in the end.
Before the wolf arrives for real, it would be better if these questions weren’t driven by someone with a stake in the outcome.
You wouldn’t ask the wolf where to keep your sheep.
