Artificial intelligence: Heads of state demand rules for human control

afp | In the run-up to the UN general debate, Chancellor Friedrich Merz (CDU) and 17 other heads of state and government signed a call for binding security rules for particularly powerful AI models. Such systems, if inadequately controlled, pose “serious risks to safety and security,” the statement said.

The signatories continue to demand that artificial intelligence (AI) must remain “under human management, supervision and control”. Companies should be required to have testing, independent audits and transparent security protocols, among other things.

As justification, the signatories point to recent cases in which powerful AI systems circumvented protection mechanisms, exploited vulnerabilities and gained unauthorized access to real computer systems. Rapid developments could outpace the ability to control emerging risks.

In addition to Merz, the 22 signatories of the declaration include the heads of state and government of Australia, Denmark, Finland, Canada, the Netherlands, Norway, Singapore and Spain. She was also joined by EU Commission President Ursula von der Leyen and the foreign ministers from Bahrain, Turkey and the United Arab Emirates. The call was initiated by Finland's President Alexander Stubb and Norway's Prime Minister Jonas Gahr Störe.

The USA, China, Great Britain, France, Japan and India are not yet among the signatories. The declaration remains open to further signatories.

The statement goes on to say that UN member states should consider creating an international institution that could set standards, make compliance with them verifiable and call states together when certain performance limits are exceeded.

The UN general debate in New York begins on Tuesday.