Commentary. The political argument over artificial intelligence is increasingly framed as a choice between progress and paralysis. That is the wrong choice. The more urgent question is whether governments can create rules that protect the public without locking useful technology behind bureaucracy.
Recent developments show why the debate has moved beyond abstract speculation. In Washington, lawmakers are considering how antitrust policy should apply to the concentrated power of leading AI companies, while the administration has discussed an oversight committee and promised to promote safety without restricting innovation. The Bank of England is also reviewing whether existing financial rules are adequate for “agentic” AI—systems capable of taking actions in payments, trading, cybersecurity and other operational settings.
Those developments point to a central editorial judgment: voluntary promises and general principles are not enough where AI systems can influence credit, employment, medical decisions, public information or financial markets. Rules should attach responsibility to the organizations deploying these systems, not leave citizens to discover failures after the damage is done.
The case for restraint
Supporters of stronger regulation have a practical argument. AI companies are building systems whose capabilities and uses can change faster than public agencies can inspect them. A model that appears harmless in a laboratory may become consequential when integrated into a bank, hospital, school or government service. In finance, for example, an autonomous tool could execute trades or disrupt operations at a speed that makes conventional after-the-fact enforcement inadequate.
Competition policy matters for the same reason. A small number of firms control much of the computing infrastructure, data, talent and distribution needed to develop advanced systems. If that concentration becomes permanent, consumers may face less choice while startups and researchers struggle to compete. Antitrust scrutiny is not hostility to technology; it is a way to prevent a vital technology from becoming dependent on a handful of private gatekeepers.
“The more urgent question is whether governments can create rules that protect the public without locking useful technology behind bureaucracy.”
The counterargument deserves weight
Opponents of expansive regulation are right about one danger: poorly designed rules can entrench the largest companies. Compliance costs that are manageable for a multinational may be fatal for a small developer, university laboratory or nonprofit. A licensing regime could also slow medical research, accessibility tools and productivity software while doing little to constrain overseas competitors.
There is a legitimate national-security concern as well. Governments do not want domestic firms to lose technological ground while rivals invest aggressively. Nor should policymakers assume that every risk can be eliminated. Complex systems will fail; the goal must be to reduce foreseeable harm and make failures traceable, not to promise a risk-free digital economy.
A workable middle course
That is why the best policy is neither a blanket pause nor a hands-off approach. Governments should regulate high-impact uses more strictly than low-risk applications, require independent testing for systems used in critical services, and mandate clear records of training, deployment and significant incidents. People affected by automated decisions should receive understandable explanations and a meaningful route to appeal.
Competition rules should target exclusionary conduct and bottlenecks rather than punish scale by itself. Public procurement can help by requiring safety audits, data protections and human oversight from vendors that sell to government. Regulators should also coordinate internationally where systems cross borders, while preserving room for responsible experimentation.
Our view is straightforward: innovation is not protected by the absence of rules. It is protected when the public can trust that powerful tools are tested, contestable and accountable. The politicians now debating AI should resist both panic and boosterism. The task is not to stop the future, but to ensure that those who build it cannot evade responsibility for its consequences.