Topics / Artificial intelligence
Should AI development be slowed or paused until safety is better understood?
The question is whether development of the most capable AI systems should be deliberately slowed or paused until their risks are better understood, made subject to testing and licensing while continuing, or allowed to proceed broadly as now. The disagreement turns on contested evidence about risks, benefits and international competition, and on differing judgements about how to weigh uncertain but potentially severe harms against delayed benefits.
Positions
- Yes: The risks of the most advanced AI systems, including possibly catastrophic ones, are poorly understood, and competition between companies and countries pushes developers to move faster than safety science can keep up. Frontier development should be slowed or paused, ideally by international agreement, until these systems can be shown to be safe.
- Regulate without slowing: The most powerful AI systems should face mandatory safety testing, licensing and incident reporting in proportion to their capabilities, but development should not be paused. Well-designed oversight can manage serious risks while keeping the benefits and avoiding the drawbacks of a halt.
- No: AI promises large benefits for science, medicine and prosperity, and a pause would delay them while handing the lead to less careful developers and rival states. The better path is continued development, with risks addressed through existing law, targeted rules for specific harms, and safety research carried out alongside progress.