Kontrollen over KI handler om statssikkerhet

The control of artificial intelligence is fundamentally an issue of state security for major powers like the United States and China. While these nations recognize similar technical risks such as cyberattacks, deepfakes, and loss of control over autonomous systems, they differ sharply on who should define and enforce the rules. In authoritarian states, security regulations easily blend with the preservation of political regimes and limits on criticism, whereas democracies require lawful authority, transparency, and accountability.

Global tech companies are shifting safety measures earlier into the development phase by pausing major training runs, increasing documentation, and proposing independent evaluators. However, voluntary industry promises remain vulnerable when safety measures conflict with market shares or speed. At the same time, heavy reliance on foreign technology creates strategic vulnerabilities for smaller nations like Norway, where foreign systems might become integrated into critical infrastructure such as hospitals, transport, and border control.

To protect national digital sovereignty and avoid strategic powerlessness, Norway should enforce four specific demands before allowing autonomous artificial intelligence systems access to critical functions. These requirements include independent evaluations prior to use, technically enforced and revocable government boundaries, complete traceability from decision bases to actions, and realistic fallback plans for continued operations if a supplier or model fails. True operational control requires clear answers about who authorized an action and under what mandate, rather than relying merely on corporate safety arguments or static management documents.