Artificial intelligence is no longer just a software industry; it is becoming a doctrine of power. On September 13, 2026, the Associated Press reported that Donald Trump downplayed calls to slow AI development, saying he did not want to cede America’s technological edge to China. Behind that sentence lies the real dilemma of the moment: should the world run faster in order to dominate, or slow down enough to avoid losing control of the race?
The debate arrives as warning signals multiply. American political leaders are discussing guardrails. Industry figures, including Dario Amodei, Sam Altman and Elon Musk according to several outlets, support some form of slowdown or stricter verification for the most powerful models. At the same time, the White House presents AI as a strategic field of competition in which the dominant ecosystem will set standards, capture economic benefits and shape international security.
The Fear of Losing to Beijing
Trump’s position follows a logic familiar to every major technological revolution: whoever slows down risks being overtaken. For Washington, AI now concentrates several issues at once. It promises massive productivity gains, reshapes cybersecurity, accelerates scientific research, transforms defense and redistributes value among companies, states and workers. In that context, slowing down can look like weakness.
China is the central argument in this vision. For several years, the United States has tightened controls on advanced chips, monitored technology transfers and accused some Chinese companies of trying to catch up with Western models through distillation or capability extraction. Anthropic published a threat report in September 2026 describing several forms of abusive use of its tools, including by China-linked actors, from surveillance to sensitive technical work. Even when contested or incomplete, such warnings feed an atmosphere of permanent competition.
Silicon Valley Discovers the Cost of Speed
What makes this moment different is that calls to slow down no longer come only from activists, academics or regulators. They also come from inside the system. When leaders of AI labs ask for audits, independent evaluators or international coordination, they implicitly acknowledge that the market alone can no longer organize safety. Competitive logic pushes every actor forward, even when all of them understand that some risks exceed the interest of any single company.
The word “slowdown,” however, remains ambiguous. Slow what exactly? The training of the largest models? Their public deployment? Agentic capabilities? Military use? Agents able to act across the internet? Without a precise definition, the debate can become political theater. Companies can advertise caution without deeply changing their pace. Governments can promise guardrails without agreeing on technical thresholds. And the public mostly sees a technology advancing faster than the explanations around it.
The Risk of Regulation Moving Too Slowly
The political difficulty is speed. Electoral cycles, congressional hearings and international treaties move slowly. Models, by contrast, progress through successive versions, infrastructure gains, data optimization and new uses distributed inside companies. Between a concern voiced in Congress and an enforceable rule, several generations of tools may already have changed the ground.
This is where AI escapes traditional categories. It is not merely a product to certify. It is becoming an action layer inside financial, industrial, media, military and administrative systems. A security flaw, misuse case or large-scale error can spread faster than an ordinary regulatory gap. The question is therefore not only whether innovation should continue. It is who takes responsibility when innovation becomes infrastructure.
The Industry Wants to Avoid the Political Wall
Major labs also have a very business-driven reason to call for guardrails: they want to preserve their legitimacy. An industry promising to transform the global economy cannot simply say “trust us.” Enterprise customers want legal certainty. Investors want to avoid a harsh regulatory backlash. Governments want to retain control over strategic uses. And citizens are beginning to connect AI to very concrete issues: jobs, energy, personal data, information, fraud and defense.
The paradox is that regulation can become a competitive advantage. Companies able to prove that their models are audited, documented, traceable and better controlled will reassure the most sensitive clients. In finance, healthcare, media, energy or defense, trust can matter more than raw performance. The next commercial battle will therefore not be fought only over the most powerful model, but over the most acceptable one.
A New Balance Must Be Invented
The acceleration camp and the caution camp are not as opposed as they pretend. Both are talking about power. The first wants to preserve America’s lead. The second wants to prevent a strategic technology from becoming uncontrollable or politically toxic. The real question is dosage: how can the industry build fast enough not to lose the competition, yet cleanly enough not to destroy the trust that allows that competition to continue?
For Europe, Africa and the rest of the world, this American debate has direct consequences. If the United States and China alone set the pace, other regions risk becoming mere adoption markets. They will use tools designed elsewhere, depend on foreign infrastructure and live under standards decided by others. AI is therefore not only a matter for California startups or Chinese ministries. It is a sovereignty question for anyone who still wants to matter in the digital economy.
The Race Can No Longer Be Blind
Trump is right about one thing: AI is a historic competition. But the leaders calling for caution are also right about the essential point: a race without rules can end up damaging the track itself. Technological dominance will not be measured only by parameters, data centers or billions raised. It will also be measured by the ability to build systems that companies, states and citizens are willing to integrate into real life.
The current moment looks less like a pause than a negotiation over speed. Nobody truly wants to leave the race. But more actors are beginning to understand that a victory achieved at the cost of a systemic accident would be a political defeat. AI is entering its imperial age: the moment when raw power must learn to live with responsibility.