The American computer scientist Cal Newport has recently named a phenomenon that many intuited, but that no one had defined precisely: doom trolling, which we could translate as interested catastrophism. According to Newport, it consists of publicly warning of the enormous risks of a technology while continuing to develop it, commercialize it and accelerate its implementation.
It is not an exercise in transparency, but a communication strategy that combines two apparently incompatible messages: “what we are building could have catastrophic consequences” and, at the same time, “we will continue building it because we have no alternative.” The result is a discourse that shifts responsibility from those who develop the technology to a supposed historical fatality.
Newport called this attitude “morally indefensible” and illustrated it with an analogy as simple as it is forceful: imagine that Ford warned that its trucks could spontaneously catch fire, but stated that it cannot stop manufacturing them because other manufacturers would continue doing so. Nobody would accept such an argument. We would expect you to recall the vehicles and fix the problem before continuing to sell them.
That is precisely what happens in the artificial intelligence industry. Big technology companies repeat the same pattern: extreme scenarios are announced (massive job destruction, existential risks or systems capable of escaping human control), but at the same time increasingly powerful models are presented stating that they have no alternative because, if they stop, other competitors would continue advancing.
It is an extraordinarily effective narrative. If the risks materialize, they can always claim that they had already warned of the danger, conveniently diluting their responsibility.
In late July, OpenAI resorted to this strategy again, claiming that one of its models had attempted to attack Hugging Face. The story, deliberately anthropomorphic, suggested that the system had decided to “rebel” against its creators. Nothing could be further from the truth. AI models have no intentions, desires, or will; They simply execute the processes for which they have been programmed. Attributing non-existent agency to them not only fuels scaremongering, but also blurs the responsibility of those who really make the decisions: the companies that design, train and deploy them.
All responsible industries report the risks of their products. The difference is in what they do afterwards. When a pharmaceutical company detects a serious adverse effect, it stops trials or withdraws the drug; When a car manufacturer discovers a structural defect, it launches a review campaign. The AI industry, on the other hand, turns the warning into a demonstration of technological power: our systems are so extraordinary that even we fear their consequences. The alert thus ceases to be a warning and becomes a powerful marketing tool.
Any transformative technology requires a rigorous assessment of its potential impacts. The problem arises when those who proclaim these risks maintain, at the same time, that they cannot modify their behavior because the technological race is inevitable. Inevitability thus becomes a convenient moral alibi.
But the history of technology demonstrates precisely the opposite. Industrial decisions are never inevitable. They are political, economic and business decisions adopted by specific people. If a company truly believes that its product could cause unprecedented workforce disruption or create systemic risks, the consistent response is to slow its rollout and bear the cost of acting prudently.
