Who benefits from the apocalypse

Describing the dystopian rural governance in India, an award-winning author, P. Sainath, titled his book "Everyone Loves a Good Drought." The spate of resignations at the global AI giant Anthropic, and the resultant debate pointing at the imminent end of humanity even before the decade turns, can be summarised in a similar-sounding title: "Everyone loves a good Apocalypse."

A researcher walks out of an AI lab and tells the world the machines may kill us all. Within a week he is on prime-time television saying it. Another one resigns, writes that the "world is in peril," and leaves to study poetry. Yet another one suggests that many of his colleagues are buying land to survive when the world comes to an end.

The instinct is to read these letters as evidence — about the danger, about the race of artificial intelligence, about how fast the thing is moving. But before you ask whether an extinction warning is true, ask who it serves.

For OpenAI, whose name means "we ship and you should use it," an employee quitting with a warning that the product could end humanity is a liability. It is bad for them. For Anthropic, the self-declared safety company, the public-benefit corporation, the exact same letter is not a scandal. It is the brand, confirmed. Our own people take the danger so seriously they walked away. That is not reputational damage. That is a testimonial.

As you pondered about the possible extinction, you probably missed the news and analysis of Anthropic's upcoming IPO, especially if you are not into stock market, shares and desires to make a quick buck. A valuation approaching a trillion dollars cannot rest on a useful chatbot. A trillion dollars rests on the belief that what you are buying is the most consequential technology in history. "This is so powerful it might kill everyone" is not a risk disclosure. It is a valuation claim. Danger is the scarcity. The more credible the extinction warning, the more reasonable the price.

Then there is the matter of what the resigned actually do next. One goes to study poetry and become, in his words, invisible. The other does the media rounds. If you sincerely believed the machine would kill all of us in a few years, this is not the behaviour of a sincere believer. You do not write a resignation letter and schedule a television appearance about the apocalypse; you would be doing something that matched the urgency. Either the belief is weaker than the letter, or the performance is the point. Both lead the same way.

None of this proves the warnings are false. Motives are rarely single; the researcher can half-believe it while the company fully exploits it. But the correct posture is not to read an extinction warning at face value when the person issuing it gets a Face the Nation segment and the company gets a stronger listing narrative. The question to ask first is the one journalism is built on: who benefits? Here the answer points straight at both of them.

This matters beyond Anthropic. We are being trained to read AI as an emergency — to accept its power as a given, its danger as proof of that power, and its builders as the only people who can be trusted to manage both. That is a convenient arrangement for the people selling the scarcity. It is a poor arrangement for everyone else, who are asked to choose between a saviour and an apocalypse and are offered no third option.

There is a third option. It has always been the same one: do not rent a black box you cannot audit. Ask who owns the thing, who controls it, and who benefits from the story being told about it. The resignation letters tell you less about the probability of extinction than they tell you about what an extinction warning is worth in September 2026.