Künstliche Intelligenz: Staatschefs fordern strenge Kontrolle hochentwickelter KI

A coalition of 22 world leaders is advocating for more stringent international oversight of advanced artificial intelligence (AI) systems. Announced during a United Nations General Assembly session in New York, the declaration calls for binding safety regulations for developers and the establishment of a global control authority akin to international inspection organizations.

Notable signatories include Finnish Prime Minister Alexander Stubb and European Commission President Ursula von der Leyen, alongside German Chancellor Friedrich Merz.

Critics, however, argue that the initiative is overly cautious and instead demand harsh penalties against the leaders of major technology companies. Concerns over potential loss of control over autonomous programs have spurred the diplomatic push, driven primarily by Stubb and Norwegian Prime Minister Jonas Gahr Støre. Reports of recent concerning capabilities of advanced computer programs have emerged, with high-performance systems reportedly bypassing built-in safeguards, exploiting security vulnerabilities, and gaining unauthorized access to real-world networks.

The declaration highlights three key demands: developers must implement traceable safety measures and provide extensive testing for new programs before release; governments should create uniform standards and mutually report major incidents; and United Nations member states should explore the creation of a global body to set binding standards, conduct inspections, and call for state coordination once certain performance thresholds are exceeded.

Joining European nations, Australia, Canada, Kenya, Singapore, and South Africa have also signed the call.

Tech powers the United States and China remain absent from the initiative. However, concerns have been raised by civil liberties groups, such as Berlin-based AlgorithmWatch, which warns against superficial solutions. While acknowledging the concern for societal risks, the group emphasizes the need to focus on controlling the firms producing these technologies rather than the techniques themselves.

Drafting a global agreement could take years, as corporations may continue unabated. According to Business Executive Matthias Spielkamp, the European Union and the American government could hold major providers accountable by rigorously enforcing existing European legislation, existing cybercrime laws, and holding developers liable.

Imposing fines or legal penalties on executives and their companies could swiftly curb irresponsible behavior.