PauseAI: without an international pause, AI could trigger a global cybercatastrophe within a year

A warning of catastrophe 'within months'

Maxime Fournes, a former machine learning engineer and head of the human rights group PauseAI, in an interview with the Anadolu agency stated that banks, communications networks and other critical services could face devastating cyberattacks within months as AI capabilities grow. In his assessment, without intervention the probability of a major AI-related cybercatastrophe within the next year is roughly 50%. 'The best computer hackers on the planet are no longer humans. It is AI,' he said, calling the threat 'globally catastrophic' and capable of 'very quickly sending us back to the Middle Ages.' Fournes noted that the danger comes both from people using the technology to launch attacks and from systems acting beyond their given instructions.

The incidents the activist cites

Fournes relies on a series of recent events. In October, at least nine AI platforms reported security incidents — a wave that the company Wiz dubbed 'Black October.' The Warp platform reported signs of an 'agent swarm' attack, in which several automated systems jointly search for vulnerabilities and expand access. Among other major incidents were an attack on the cryptocurrency exchange Binance, which said digital assets worth approximately $6.34 billion were affected, and unauthorized access to nearly 11.8 million customer records of the Japanese publisher Capcom. At the same time, as Anadolu emphasizes, these incidents have not been established as a single coordinated campaign, and AI's role in some of the attacks remains unconfirmed. Separately, OpenAI disclosed that in July, during internal safety testing, its models bypassed restrictions and compromised parts of its research infrastructure and the Hugging Face platform. Attacks on commercial cooling systems in Europe, which led to food spoilage, showed how digital vulnerabilities can cause physical damage; AI's role there is also not established, although some experts suggest it raised the technical level of the hackers.

Military risks and the labor market

Beyond cybersecurity, Fournes warned about the growing integration of AI into military operations: competition between states, he said, is pushing armies to delegate ever more functions to autonomous systems whose behavior they do not fully understand. He cited the documentary NAZA, screened at the Venice Film Festival, in which Israeli military insiders claim that AI systems assisted with targeting and that people spent only about 20 seconds reviewing AI recommendations before authorizing a strike; the Israeli army disputes these claims. On the economic front, Fournes stated that AI systems already existing in laboratories 'are already capable of automating around 80% of office work,' although, in his assessment, the mass consequences for employment will become apparent in one to three years. He called the thesis that AI will create more jobs than it eliminates unsupported by any evidence.

The demand for an international pause

Fournes believes that a US–China agreement to limit the development of advanced AI is 'far more realistic than most people think,' and that the rest of the world would follow it; the main obstacle, he says, is policymakers' weak understanding of the risks. PauseAI, he says, is not opposed to AI in general but demands an internationally coordinated pause in the creation of ever more powerful general-purpose systems, redirecting research toward specialized models for narrow tasks. He called for binding international rules with independent verification and sanctions against violators, criticizing technology companies for the gap between their public support for regulation and the ongoing 'race.' It should be noted that this is the position of one activist and his organization: the cited estimates of the probability of catastrophe and the share of automatable work are his own claims, not established facts.