As fears over artificial intelligence escalate, U.S. policymakers are increasingly drawn to regulatory frameworks that would deliberately slow the pace of technological advancement. Industry leaders have signaled support for such caution, but a growing number of analysts warn that the cure could be worse than the disease.
At the heart of the debate is a fundamental tension: enforcing an AI slowdown would demand an extraordinary expansion of government surveillance and investigative powers. History suggests that the risks of such concentrated authority may outweigh the dangers of unchecked development.
Targeted Rules vs. Broad Restraints
Governments can effectively regulate specific uses of AI—banning fraud, safeguarding privacy, upholding copyright, and assigning liability for harm. The real challenge emerges when regulation aims to halt technological progress itself, prohibiting private firms from developing capabilities deemed too dangerous.
Any such restraint faces a basic obstacle: the companies targeted would have powerful incentives to evade it. The profit motive, which fuels innovation, would also drive firms to secretly develop prohibited technologies. Unlike a factory, which cannot easily hide its existence, an AI firm could operate with superior algorithms and data without immediately revealing its edge.
Monitoring the Unmonitorable
To enforce a slowdown, regulators would need deep access to corporate R&D—tracking computing resources, training runs, data centers, internal evaluations, and research projects. They would have to distinguish routine improvements from breakthroughs that cross arbitrary lines. Given the rapid evolution of frontier AI, this oversight would need to be continuous, not occasional.
The scale of such monitoring would require a government agency with extraordinary technical expertise and access to the most commercially sensitive information in the economy. That concentration of knowledge would invite conflicts of interest, regulatory capture, and corruption—a political economy dimension often overlooked in the AI safety debate.
Authoritarian Precedents
History offers cautionary tales. Many authoritarian states have justified expansive state power by promising safety and security, often at enormous societal cost. The question becomes: How powerful would a government have to become to guarantee that no one is secretly developing prohibited AI?
Internationally, the problem grows even more acute. A U.S. slowdown would be futile if other technological powers continue racing ahead. Arms control treaties from the Cold War provide some precedent, but they relied on observable physical artifacts—missiles, launchers, nuclear tests. AI progress is far harder to monitor. Would China accept comparable American oversight of its technological infrastructure?
This leaves Washington with an unpalatable choice: adhere to restrictions that rivals might evade, risking technological decline, or abandon its own rules upon suspicion of cheating. Neither option builds a durable regulatory regime. As public confidence in the federal government hits historic lows, the legitimacy of such expansive oversight is questionable.
A New Concentration of Power
None of this suggests AI is safe or that governments should do nothing. Specific applications will inevitably be regulated. But the push to slow technological progress—as seen in the backlash against AI deregulation—demands a broader conversation. The usual safety debate asks how much power we leave with private companies; it must also ask how much power we are willing to grant the state to stop them.
Creating a government powerful enough to monitor the frontier of innovation is not a costless response to AI risks. It introduces a new concentration of power—and a new risk of its own. As rifts over AI regulation emerge within the administration, these trade-offs deserve urgent scrutiny.
