Commentary. Artificial intelligence is moving faster than the institutions meant to govern it. That familiar observation has acquired sharper force as the European Union’s AI Act becomes enforceable, the United States continues to rely heavily on voluntary commitments and executive action, and China pursues a more state-directed model of oversight. The disagreement is not whether AI should be governed. It is who should set the limits, how binding those limits should be, and who pays when systems fail.
Our view is straightforward: governments should establish enforceable baseline protections, while leaving room for responsible experimentation. Neither a regulatory vacuum nor a blanket ban is an adequate answer to technologies already influencing hiring, education, finance, healthcare and public services.
The case for enforceable guardrails
The strongest argument for binding rules is practical rather than ideological. Automated systems can affect access to jobs, housing, credit and education, often through decisions that are difficult for an ordinary person to understand or challenge. The Center for Democracy and Technology reports that 2026 legislation has increasingly addressed automated decision-making in these consequential areas, as well as frontier-model risks and independent auditing.[1]
These are not abstract concerns. If a person is rejected for a loan or screened out of employment, an explanation after the fact may be worthless unless there is also a route to correction. Rules requiring documentation, testing, human oversight and meaningful appeal would not eliminate discrimination or error. They would, however, make powerful institutions more accountable.
The EU has chosen the clearest version of this approach. Its AI Act became enforceable on Aug. 2, 2026, according to legal analysis from Simmons & Simmons.[2] Europe’s framework offers a public benchmark: companies can know in advance which obligations apply, and citizens can point to a formal rights-based system rather than relying solely on corporate promises.
The objections deserve an answer
Critics are right to warn that regulation can impose costs that fall hardest on smaller companies. Compliance paperwork, audits and legal uncertainty may entrench the largest technology firms, which can afford teams of lawyers and engineers. Europe has already softened elements of its approach under industry pressure, Euronews reported, illustrating the tension between ambition and implementation.[3]
There is also a legitimate innovation argument. A start-up developing a medical tool or an accessibility service should not face the same burdens as a company deploying a large system across millions of people. Rules that treat every algorithm as equally dangerous would be both inefficient and politically fragile.
But these objections argue for proportionate regulation, not no regulation. A risk-based system can exempt low-impact uses, offer practical guidance to smaller firms and create testing environments in which new products can be evaluated before broad deployment. The alternative—asking the public to absorb the costs of mistakes while companies retain the gains—is not neutrality. It is a policy choice.
A fragmented global settlement
The emerging international split makes coordination harder. Euronews describes the EU as relying on binding laws, the United States on voluntary promises and executive orders, and China on a more centralized model.[3] None is a perfect template. Europe risks overregulation; the United States risks uneven protection; China’s model raises obvious concerns about state control and transparency.
Still, the differences should not become an excuse for inaction. Governments should agree on a narrow common floor: no secret high-impact decision without human accountability, no deployment without documented risk assessment, strong protection for personal data, and rapid incident reporting when systems cause serious harm. Such standards would preserve national choices while limiting regulatory arbitrage.
At Archange Shadows, we do not believe every AI question can be settled by legislation. Professional ethics, technical standards, journalism and informed users all matter. Nor should policymakers pretend that existing institutions can simply be copied into an unfamiliar technological environment.
Yet the burden of proof should rest with those deploying systems capable of changing people’s lives at scale. Innovation remains valuable, but speed is not a public virtue when accountability is absent. The durable bargain is neither “move fast and break things” nor “stop everything.” It is permission to build, matched by a duty to explain, test and repair.