“It was exactly one year ago that Nobel Peace Prize laureate Maria Ressa stood at the United Nations and launched our Global Call for AI Red Lines. The year since has only confirmed our worst fears about the dangers advanced AI poses to our societies when it is developed without any binding international framework. We urgently need shared rules and institutions, clear red lines that must never be crossed, and independent oversight mechanisms. This new declaration is a significant step in that direction, and we call on governments around the world to join the process.”
A shared diagnosis: the most advanced AI systems must remain under human control
The declaration brings together governments from several regions of the world, including Australia, Canada, Denmark, Finland, Germany, Ireland, the Netherlands, Norway, Singapore, Spain and the United Arab Emirates, as well as European Commission President Ursula von der Leyen. It warns that frontier AI models are advancing so quickly that the risks they create are growing faster than our societies can manage them.
The signatories are therefore calling on:
- companies to carry out pre-deployment testing and submit to independent evaluations;
- governments to adopt common standards and report serious incidents;
- the UN to consider an international body able to set standards, verify compliance and convene states when critical thresholds are crossed.
CeSIA shares this assessment and welcomes the statement, which follows directly in the spirit of the Global Call for AI Red Lines, launched a year earlier by twelve Nobel laureates and hundreds of prominent figures.
CeSIA calls on other governments, especially democracies around the world and France in particular, to join this effort and help build international governance of artificial intelligence.
In recent weeks, a rapid succession of AI-related harms and incidents of growing severity has shown how urgently we need shared guardrails and protections against these risks. After Hugging Face and OpenAI were covertly hacked by hundreds of agents belonging to OpenAI itself, Anthropic reported that it had detected several deeply concerning uses, and attempted uses, of its own systems, notably in cybersecurity and biological weapons.