Norway’s artificial intelligence policy assumes that human oversight ensures safety and accountability in AI systems. Yet a recent study indicates this assumption is flawed. Researchers at the University of Oslo found that human control frequently fails to prevent errors or biases in AI applications. The gap between policy and practice raises concerns about the country’s readiness for advanced AI deployment.
The study examined AI systems used in public sector decision-making, including welfare assessments and healthcare diagnostics. In multiple cases, human reviewers either ignored AI recommendations or lacked the technical expertise to challenge them. One case involved an AI tool that incorrectly denied social benefits to 15% of applicants due to flawed data. Officials admitted oversight was minimal because reviewers were not trained to question algorithmic decisions.
Norway’s approach contrasts with Singapore’s regulatory framework, which mandates independent audits and real-time monitoring of AI systems. Singapore’s model requires developers to document data sources and decision logic, making it easier to detect flaws. Norwegian officials acknowledge the need for stricter rules but cite limited resources as a barrier.
The debate extends beyond Norway. The European Union’s AI Act, set to take effect next year, emphasizes human oversight but leaves implementation details to member states. Critics argue this vagueness could lead to inconsistent enforcement. The Norwegian Data Protection Authority has called for clearer guidelines but has not proposed specific changes.
For now, Norway’s reliance on human oversight remains a weak link in its AI strategy. Without stronger safeguards, the country risks repeating the mistakes seen in other nations where AI systems caused real harm despite supposed human checks.
Source: digi.no