Drafting the EU’s AI Act felt “like writing traffic laws while the cars were already crashing,” an EU lawmaker recently revealed in an off-the-record confession. I couldn’t shake that picture. It’s painfully accurate in addition to being evocative.
Governments are rushing to contain artificial intelligence rather than merely responding to it. The way they’re constructing new legal frameworks around generative AI is the best example of this urgency.
| Issue | Description |
|---|---|
| Core Concern | Generative AI is evolving faster than legal systems can respond |
| Urgency Drivers | Misinformation, bias, deepfakes, copyright infringement, lack of accountability |
| Governmental Actions | EU AI Act, Biden Executive Order, UK AI Safety Institute, China’s provisional AI measures |
| Key Policy Goals | Transparency, safety, fairness, intellectual property protection, international cooperation |
| Strategic Balance | Foster innovation while preventing societal and economic harm |
ChatGPT was more than just a novelty when it debuted in late 2022; it was a shock. Students were suddenly producing essays in a matter of seconds. Coders were using prompts to debug. And con artists? They had access to devices that could remarkably easily create identities, mimic voices, and replicate faces.
A flurry of policy initiatives ensued within a few months. With a focus on foundation model safeguards, the European Union hurried to complete the AI Act. In response, the US issued an executive order calling for watermarking, safety testing, and transparency. The UK established its AI Safety Institute in the meantime, establishing itself as a pioneer in responsible development.
The theme of these measures is remarkably similar: lawmakers want to guide AI, not stop it. And quickly.
However, why the abrupt worldwide convergence? What AI is already upending holds part of the solution. Consider false information. Deepfakes, especially during election cycles, have evolved from a fringe technology to a serious threat. At least two nations issued emergency briefings in response to the generative tool-created, viral audio hoaxes that went viral in Latin America last year.
Then there’s the copyright problem. Large datasets, including writing, music, and images, are frequently scraped by generative models without authorization. Lawsuits have been filed by publishers and artists, some of whom claim that opaque algorithmic engines are stealing their jobs. Legislators are being compelled by these disputes to reconsider outdated definitions of intellectual property and pose fresh queries regarding authorship, ownership, and just compensation.
Safety issues, on the other hand, tend to feel even more nebulous—but potentially more disastrous. Consider incorrect medical diagnoses, malfunctioning autonomous vehicles, or criminal justice systems that use biased sentencing guidelines. When AI makes a mistake, who is responsible? The developer, is that right? The person who deploys? The user?
Because of this ambiguity, “accountability” is now the most contentious term in AI policy circles. And with good cause. Innovation can slip into impunity in the absence of clear lines of accountability.
A policy advisor from East Africa noted that these gaps are not only legal but also economic during a closed-door meeting in Geneva last spring. Without regulation, biased systems will be imported. We run the risk of impeding local startups if we regulate too soon,” she said. A large portion of policymaking in developing nations is motivated by this conflict between promoting domestic innovation and safeguarding citizens.
This push-pull dynamic is particularly evident in the US. Many states are experimenting with their own laws while the federal government increases its investment in AI. Biometric privacy clauses are being investigated in California. Texas is concentrating on school safety equipment. Election-related deepfakes have been targeted by New York. Although it’s a patchwork of regulations, it shows that both parties agree that AI needs safeguards.
I recall reading a footnote in a Brookings report about how, in the first half of 2025 alone, 47 U.S. states passed legislation pertaining to artificial intelligence. Coordination is not what that is. That’s alarm-driven momentum.
The EU is trying something more centralized on the other side of the Atlantic. The AI Act’s tiered risk model, which places strict oversight on high-risk tools and lax regulation on low-risk ones, is especially innovative. This includes strict penalties for non-compliance, transparency requirements, and documentation mandates. Notably, even if generative models are developed outside of the EU, the Act mandates that they comply with EU copyright law.
In contrast, China has tended to use command-and-control strategies. Provisional regulations requiring businesses to register and pre-approve large-scale generative tools have already been put into place. Beijing’s strategy combines an attempt to convey confidence in its technological governance with a national security perspective. It is domestically tailored, strictly regulated, and top-down.
Nevertheless, despite these differences, one idea keeps coming up: AI is too significant to be left to chance. and too strong to be left unchecked.
“AI governance must be anticipatory, inclusive, and enforceable,” according to a G7 communique. It’s not as simple as it seems. The technical know-how to comprehend, let alone control, quickly developing AI systems is still lacking in many governments. Additionally, even though task forces and advisory bodies are forming at an unprecedented rate, their recommendations frequently fall short of the capabilities of the instruments they are intended to restrict.
Nevertheless, this regulatory sprint has a notable sense of optimism. Lawmakers have been lagging behind technology for decades, from cybersecurity to privacy. However, there is a deliberate effort to advance, establish standards early, and develop cross-border interoperable frameworks with generative AI.
Governments are doing more than just reducing risk when they harmonize policies across regions. For safety, they are creating a common language. In the digital age, where a voice-cloning app developed in one nation can trick a user in another before breakfast, this is very important.
Additionally, there is a strategic interest involved. AI is a significant advancement in both technology and the economy. Properly crafted regulations can protect citizens and spur domestic innovation in a country. Those who don’t run the risk of falling behind or, worse, importing uncontrollable systems.
One instance from last year caught my attention: a UN advisor acknowledged, “We used to regulate after the crisis,” during a conference coffee break. We’re attempting to stop one with AI. That mental change, no matter how small, might turn out to be the most important one of all.
Therefore, even though the frameworks are still developing and the definitions are still up for debate, there is a clear movement in progress. Governments are not merely rushing to control AI. In a time when machines are increasingly writing, speaking, and making decisions alongside humans, they are vying for the public’s trust.
