Parliaments must be central to the democratic governance of advanced artificial intelligence (AI), according to a new publication from the Inter-Parliamentary Union (IPU), including helping to address the potentially catastrophic risks posed by the technology.
Democratic control of artificial intelligence: The role of parliaments in addressing extreme risks identifies practical steps lawmakers can take to strengthen democratic accountability, international cooperation and support public safety in the face of the rapid pace of AI development and deployment.
The IPU brief examines four main extreme AI risks: large-scale cyberattacks, biological and chemical threats, manipulation of democratic processes and the potential loss of human control over highly autonomous systems.
It also considers overlapping risks arising from military applications, including the use of AI in targeting nuclear decision-support systems.
The brief says recent incidents have exposed weaknesses in safeguards. It highlights the widely reported story of Hugging Face, an AI platform in which AI agents bypassed human controls and attacked other online platforms and government systems.
Such incidents have brought extreme AI risks onto the agenda of many parliaments.
The brief identifies ten entry points for parliamentary action, including national risk assessments, safety requirements for advanced AI entering national markets, clearer legal liability and stronger requirements to report serious incidents.
All parliaments can act, including countries that mainly import AI systems by influencing domestic standards through purchasing rules, market access conditions and international cooperation.
Protecting democratic processes is a particular focus of the publication. Parliaments should work with electoral authorities, political parties, civil society and experts to examine the wider impact of AI on democratic life and raise public awareness of risks such as misinformation and electoral manipulation.
The brief highlights the Global Call for AI Red Lines, a multi-partner initiative promoting international agreement on unacceptable uses and behaviours of AI.
The IPU publication also showcases AI action taken by parliaments and the work of their members through lawmaking, scrutiny of government actions and public engagement.
The new brief forms part of the IPU’s wider work on artificial intelligence. The IPU supports parliaments through multi-stakeholder dialogues, exchanges between parliamentarians, and the sharing of experience and good practice on both the use of AI within parliaments and the development of AI policy around the world.
