Governments around the world are facing growing difficulty regulating artificial intelligence systems whose capabilities and deployment models evolve faster than existing legislative processes.
In the European Union, implementation of the AI Act has begun, focusing on risk-based classification and compliance requirements. In the United States, regulatory authority remains fragmented across agencies, with enforcement relying largely on existing consumer protection and civil rights laws.
Policymakers acknowledge that traditional regulatory approaches are ill-suited to technologies that update continuously and are deployed across borders.
“By the time a rule is written, the underlying system has already changed,” said one government advisor involved in AI policy. “We’re regulating moving targets.”
One challenge is scope. Modern AI systems are general-purpose, meaning the same model can be used for benign tasks such as summarization and for high-stakes applications such as surveillance or automated decision-making. Regulating by use case has proven complex in practice.
Another issue is transparency. Companies often cite intellectual property concerns when asked to disclose training data, model architecture, or evaluation methods. This limits regulators’ ability to assess risks independently.
As a result, some governments are shifting toward process-based oversight, emphasizing auditing, documentation, and accountability structures rather than prescriptive technical rules. Others are exploring public-private partnerships to improve regulatory capacity.
Critics argue that these approaches still lag behind industry deployment and may entrench large firms that can afford compliance costs, further concentrating power.
Despite these concerns, few policymakers advocate slowing AI development outright. Most see AI as economically and strategically essential, even as governance gaps widen.
For now, regulation remains reactive. Whether it can become anticipatory without stifling innovation is an open question — one that will shape how AI systems are integrated into public life over the coming decade.
