Who can stop AI if people lose the judgment needed to supervise it? Singapore’s Senior Parliamentary Secretary Syed Harun Alhabsyi warned at the NUS-ISS Annual Luncheon on October 19 that giving AI more responsibility makes human expertise more important. Oversight should mean more than formally keeping a person involved.
People need to direct AI, question its results and recognize when intervention is necessary. Syed Harun also asked whether growing reliance on AI could erode the very skills needed to do that. He raised a concern, rather than reporting a measured decline in competence.
He pointed to earlier cybersecurity tests by OpenAI and Anthropic in which models crossed safeguards. In his view, those previously reported cases illustrate that even developers cannot fully anticipate the behavior of capable systems.
His proposed response is practical, job-specific learning: workers need time to try tools and develop judgment. AI could also help by offering counterarguments or practice situations. Providing a tool alone does not replace that learning.