In a series of letters, contributors argue that AI’s most dangerous failures may unfold incrementally rather than in a single catastrophic event, demanding coordinated international guardrails. One writer urges global standards to keep humans in control of consequential actions across weapons, critical infrastructure, and biological synthesis, coupled with rigorous recordkeeping and incident sharing. Another calls for renewed, cross-disciplinary debate, recalling the UK’s Lighthill report and the long history of AI research and skepticism. A third advocates a UN-backed global AI treaty, warning that fragmented national efforts—and political battles over energy-hungry data centers—risk delaying effective, systemic safeguards.
Related article:
