AI Is Now Moving Faster Than Governments Can Govern It

In the time it takes a government to draft a single AI policy, a new generation of models has already been released—and rendered it obsolete. That is the stark reality facing regulators worldwide as artificial intelligence accelerates faster than any legal, ethical, or safety framework can keep pace. The core problem is simple: governments lack the technical expertise, institutional speed, and adaptive structures needed to evaluate and regulate advanced AI systems, forcing them into a cycle of reactive policymaking that leaves societies exposed to mounting risks. This AI regulation lag is not a future concern—it is happening now, and it is widening every quarter.

The headline from Forbes' Craig Smith captures the essence: 'AI Is Now Moving Faster Than Governments Can Govern It.' The statement reflects a consensus among technologists, policymakers, and academics who have watched the gap between innovation and governance grow from a crack into a chasm. While the European Union has taken the most ambitious step with its AI Act, the law is still being phased in, and its provisions may already be outdated by the time they fully take effect. The United States has relied on executive orders and voluntary commitments from companies, but no comprehensive federal AI regulation exists. China has imposed strict rules on recommendation algorithms and deepfakes, yet its state-backed AI companies continue to push boundaries.

Why now? The release of GPT-4 in early 2023 marked a turning point—it was the moment the general public and governments alike realised that AI was no longer a research curiosity but a transformative force. Since then, each subsequent model—Claude 3, Gemini 1.5, and others—has brought capabilities that were unthinkable a year prior. The speed of improvement has been exponential, while the speed of governance remains linear at best. According to a 2025 analysis by the Center for AI Safety, the average legislative process for technology regulation in advanced economies takes 3 to 5 years, while AI model capability doubles roughly every 18 months. That mismatch means any law passed today is aimed at a target that has already moved.

The consequences of this AI regulation lag are already visible. In financial markets, AI trading bots have triggered flash crashes that regulators could not explain in real time. In healthcare, unapproved AI diagnostic tools are quietly being used by hospitals. In democratic processes, deepfakes and AI-generated disinformation have influenced elections in at least four countries since 2024. Governments respond after the fact—banning a specific tool, issuing a fine, or launching an investigation—but they are always behind the next innovation. The reactive policymaking approach creates a whack-a-mole dynamic that favours the fastest movers in the private sector.

What this means is that the burden of responsible AI development is shifting onto the companies themselves. Several leading AI labs have established internal safety boards and pledged to implement measures like 'compute thresholds'—caps on the amount of computing power used for training dangerous models. But these are voluntary and unevenly enforced. Critics argue that self-regulation is no substitute for binding rules, especially when market pressures incentivise speed over safety. Informed observers point to a need for entirely new models of governance—something like a 'rapid response' regulatory agency that can issue binding guidance in weeks, not years. Some have proposed a global AI watchdog analogous to the International Atomic Energy Agency, but political will remains elusive.

Looking ahead, the next 12 to 18 months will be decisive. The EU AI Act's full implementation in 2027 will provide a real-world test of whether structured regulation can adapt. The US Congress is debating a handful of bills, but none have passed both chambers. Meanwhile, the next generation of AI models—some reportedly capable of autonomous software engineering—is expected by mid-2026. If governments cannot accelerate their own processes, the AI regulation lag will only deepen. The question no longer is whether AI needs governance, but whether governance can reinvent itself fast enough to matter. The answer will shape not just the future of technology, but the safety and agency of every person on the planet.