A group of scientists serving on a United Nations artificial-intelligence committee cautioned against the use of catastrophic language when discussing the technology’s dangers. The warning was delivered during a side event in New York that coincided with the UN General Assembly. Panel members argue that alarmist narratives can distract from rigorous research and policy work needed to understand and manage AI systems.
Joelle Barral, an executive at Google DeepMind and a member of the International Scientific Panel on AI, emphasized the need for scientific investment over fear-driven narratives. “As scientists, it’s very important to invest in the science, to really understand what we know (from) what we don’t know, and investing in fear is not that helpful,” she said. Barral urged colleagues to focus on empirical gaps rather than speculative doom.
Former Anthropic employee Jacob Coxon, who previously worked at OpenAI, warned in early September that developers sometimes claim the technology “could kill us all by the end of the decade.” A research engineer who left Google DeepMind in July echoed a similar sentiment, suggesting it might already be “too late to avoid” humanity’s eradication by AI. Both remarks illustrate the extreme scenarios circulating among some technologists.
Nobel Peace Prize laureate Maria Ressa, an investigative reporter, called for a shift in how AI hazards are framed. “We have to change the way we talk about AI risk because for a very long time, it’s either existential and everything is going to die and the killer robots are coming, or it doesn’t really matter,” she told the audience at the New York event on September 21. Ressa’s appeal highlights the need for balanced discourse.
The International Scientific Panel on AI was established in 2025, with its membership finalized in February 2026. Its charter tasks the body with deepening scientific comprehension of artificial intelligence and feeding that knowledge into United Nations deliberations. Co-chair Yoshua Bengio, widely regarded as a founding figure of modern AI, warned that it is crucial to separate evidence-backed risks from what can only be “plausibly extrapolated” based on current data. The panel aims to provide that distinction.
Collectively, the panel’s members urge policymakers, developers and the public to move beyond sensational headlines and concentrate on measurable uncertainties. By grounding discussions in verified research, they hope to shape regulations that address genuine threats without stifling beneficial innovation. The call for restraint reflects a broader effort within the AI community to align rhetoric with scientific reality.