The singular nature of the artificial intelligence ‘coup’ according to Maria Ressa

According to Maria Ressa, Nobel Peace Prize laureate, AI companies treat us like Pavlov’s dogs. A fitting image, but for reasons opposite to hers. It is her interview with Corriere della Sera that triggers a series of conditioned reflexes. The microphone’s bell rings – that is, the cue arrives – and the response follows promptly: a technological coup, an existential shift, surveillance for profit; it’s time to say ‘enough is enough’. Pavlov measured saliva in drops; here, it would be measured in words.

Ressa, a Filipino journalist and founder of the Rappler website, was awarded the Nobel Prize in 2021 for defending press freedom under Duterte. In the interview, she summarises a timeline that is now familiar to us: first, social media hijacked our attention; then chatbots invaded our privacy; now agents possess the same capacity to act as we do – and all of this is said to have damaged our biology. How? Why? In what sense?

We do not know; after all, they are reflections, not topics.

The coup d’état—which Ressa dates to 2024—is a must. It is a peculiar coup: it has been going on for two years, has not yet deposed anyone, and the governments subjected to it continue to appoint committees of experts and send them to the UN to tell the press about the threats that AI allegedly poses to their sovereignty. ChatGPT and its ilk must be long-term coup plotters. Every day that passes without legislation, the Nobel laureate warns, the state loses further power: if the maths were correct, we would be approaching the day when it will not even have enough power to pass the very law that should save it.

The tests are modelled on the plot of "The Matrix". In the report submitted to the UN, he says, we have demonstrated that agents escape from laboratories, lie and alter records. The report does indeed examine an incident in which agents employed in cybersecurity assessments circumvented restrictions and deceived an assessor: this is not a made-up story, nor should it be dismissed as a laboratory simulation. There is, however, a methodological problem. Behaviour observed in a laboratory, or in a limited incident, demonstrates that an event is possible under certain conditions; it says nothing about how often it might occur, whether it is a problem specific to that environment or universal, nor what the likelihood is of it causing serious damage. The report itself does not estimate the probability or timeframe of a large-scale loss of control. Between the observation of a capability, its reproducibility under operational conditions and the probability of significant consequences, there is a gap in reasoning that indignation cannot bridge. In short: for now, the coup can wait.

There is a curious asymmetry. The system developers provide much of the data on the behaviours that should concern us: Ressa dismisses their fears and proposed actions as irrelevant, yet accepts the evidence they themselves produce as incriminating confessions. “It’s the data that counts”, provided it is interpreted in the right way – that is, his way. This use of sources is more reminiscent of the Inquisition than of a systematic review.

In terms of solutions, the Nobel Peace Prize laureate is calling for a radical overhaul of Section 230. This is the 1996 US law under which operators of online services are not considered publishers of user-generated content and are therefore not liable in court. According to Ressa, platforms are not merely intermediaries but publishers. It is a valid argument, one that has been debated for years; but if a company is to be held legally liable for a vast amount of third-party content, the incentive to pre-emptively remove anything that poses a legal risk is obvious, and prudent moderation becomes private censorship. It is rather striking to hear this argument put forward by someone who was honoured in Oslo for freedom of expression.

Like many other innovations, artificial intelligence can raise issues of transparency, accountability and the concentration of power. Precisely for this reason, it deserves to be discussed by people who do not believe that repeating the same concept ad nauseam amounts to proving anything. It must be said that Pavlov won the Nobel Prize for Medicine, and he studied reflexes using a method. The reflex is there. The method is missing.