UN Pushes Global AI Treaty for Human Control as Risks Spread to 3 Critical Sectors
Updated
Updated · The Guardian · Aug 30
UN Pushes Global AI Treaty for Human Control as Risks Spread to 3 Critical Sectors
2 articles · Updated · The Guardian · Aug 30
Summary
An international effort is seeking broad political backing for a UN-backed AI treaty after last month’s UN Global Dialogue on AI Governance in Geneva called for human control over advanced systems.
Three areas are cited as immediate red lines—weapons systems, critical infrastructure and biological synthesis—where advocates want minimum safeguards, clear human authorization, audit records and international reporting of failures and near-misses.
The push argues AI’s gravest harms may emerge without a dramatic “takeover,” including helping design pathogens, exposing infrastructure vulnerabilities or quietly expanding autonomy through many routine decisions.
Supporters say fragmented national or corporate approaches will delay effective controls as AI development accelerates, while energy-hungry datacentre disputes are also diverting political attention from broader governance.
The appeal casts AI as a global systemic risk akin to climate change, reviving a debate that some contributors note has existed since at least 1972 but now carries greater urgency because computing power has caught up.