Doom trolling is what happens when PR stops respecting the audience

Gabriel De La Rosa Cols is venture partner and head of communications at Add Ventures Capital. 

Jacob Coxon recently resigned from Anthropic and published the kind of warning that, a few years ago, would have sounded extraordinary. After three years working at OpenAI and Anthropic, he accused both companies of “racing straight to self-improving superintelligence and gambling with our lives.” Then came the line that caught my attention: “This is not a marketing stunt.”

I don’t believe him. Or, more precisely, I no longer know why I should.

That is not entirely Coxon’s fault. It is the result of a communications strategy the AI industry has spent years perfecting: tell the public your product may destroy their job, destabilize society and, if the day is slow enough, kill everyone. Then act surprised when the public suspects the apocalypse is part of the media plan.

Cal Newport has a good name for this: doom trolling. He uses it to describe the bizarre habit of AI companies trying to convince customers that the products they are building may eventually cause massive harm. It is hype through pessimism. Instead of saying your product will change the world, you say it may end it.

The logic is simple. If AI can eliminate millions of jobs, it must be incredibly powerful. If it can destabilize civilization, the company building it must be incredibly important. If it can potentially wipe out humanity, imagine what it can do for your Excel spreadsheet.

For communications people, the problem should be obvious. Credibility is not infinite. Every grandiose claim is a withdrawal. Eventually the account is empty.

Then someone like Coxon comes along and says, sincerely or otherwise, that we should be terrified. The audience has already been trained to hear a sales pitch.

Ed Zitron, a tech critic who also works in PR, reacted to Coxon with a wonderfully cynical line: “the biggest sign that something is a marketing stunt is when someone says that.” Funny, yes. But also devastating. “This is not marketing” only becomes necessary after marketing has made truth indistinguishable from a stunt.

You can see the same self-mythologizing in the way Sam Altman talks about OpenAI. The New Yorker noted this year that Altman has repeatedly compared himself and OpenAI’s mission to Oppenheimer and the Manhattan Project. At some point the analogy stops sounding historical and starts looking narcissistic.

There is something almost touching about the self-importance required to look at software that occasionally invents court cases and conclude that history has placed you in the moral universe of the atomic bomb.

But the mythology serves a purpose. Useful software is not enough. Useful is boring. Useful does not justify breathless headlines or trillion-dollar expectations. The technology has to be inevitable. Transformative. Civilizational. Possibly lethal.

The same formula powers the “jobpocalypse.” Anthropic CEO Dario Amodei warned that AI could wipe out half of entry-level white-collar jobs within one to five years and push unemployment to 10% to 20%. Of course that generated headlines. “New software may improve productivity” is terrible copy. “Your children may never have jobs” is much better.

This is where communicators should ask an uncomfortable question: what are we optimizing for?

If the answer is attention, doom trolling works brilliantly. If the answer is trust, it is a disaster.

The first warning scares people. The tenth makes them skeptical. The hundredth makes them think, just kill us already.

People are tired of being treated like useful idiots in somebody else’s hype cycle. We are expected to believe these models hallucinate basic facts and fail in stupid ways, but are also on the verge of replacing half the economy and acquiring godlike intelligence.

Maybe both things will eventually be true. But extraordinary claims require more than another founder interview, safety report or viral resignation post.

And no, dangerous does not equal sexy in the real world, at least not to mature adults. That is an infantile idea. If a bank bragged that its new risk engine might collapse the global economy, nobody would think: impressive technology. They would move their money.

This is where doom trolling becomes a PR industry problem.

Public relations has spent decades trying to escape the caricature of the spin doctor: the professional bullshitter paid to make clients look better than reality warrants. Then one of the most important industries on earth adopts a communications strategy that seems to say: yes, we are manipulating you, but we replaced optimism with existential dread, so technically it is innovative.

We should be the people in the room saying no.

Not every benchmark is a historic turning point. Speculative risk is not a product benefit. The CEO does not need to be Oppenheimer. Frightening people is not the same thing as persuading them.

AI companies already have a great story. LLMs can write, code, summarize, translate, analyze information and automate parts of knowledge work at astonishing speed. They are also unreliable, prone to hallucinations and dependent on human judgment.

That is not a weak story. It is an honest one.

The job of PR is not to manufacture importance. It is to recognize what is genuinely important and explain it convincingly. That means treating audiences as adults: saying what a technology can do, what it cannot do, where risks are demonstrated and where everyone is still speculating.

The moment audiences believe you are spinning them, they begin discounting everything you say, including the things that are true.

Coxon may be sincere. That sincerity itself now sounds like marketing should terrify AI companies more than another Terminator scenario.

The best PR builds trust through honesty, empathy and respect for the audience’s intelligence. If the truth is not exciting enough, that is a messaging problem, not an excuse to manufacture an apocalypse.