AI Regulation: How the US, China and Europe Are Taking Different Paths

Artificial intelligence is advancing faster than governments can establish common rules for its development and deployment. The United States, China and the European Union are now pursuing markedly different regulatory strategies, revealing competing views on how to balance innovation, national interests and the risks posed by increasingly powerful AI systems.

While Washington is leaning toward voluntary commitments and legal responsibility after harm occurs, Beijing is combining strict oversight of applications with continued support for technological development. The European Union, meanwhile, has chosen a rules-based framework designed to impose obligations according to the level of risk posed by AI systems.

United States: Innovation First, With Limited Federal Oversight

Under President Donald Trump, the United States has largely favored an approach designed to avoid regulatory constraints that could slow the country's AI industry.

The federal government has relied heavily on executive action and voluntary commitments from technology companies rather than introducing a comprehensive law requiring independent safety assessments before the most advanced models are released.

Trump also reversed a 2023 policy introduced under Joe Biden that had required developers of powerful AI systems to provide the government with information about their safety testing.

A subsequent executive measure created an avenue for companies to voluntarily submit certain highly capable systems for government review before launch, particularly models considered capable of conducting large-scale cyberattacks.

The debate has intensified following tests in which AI systems developed by major technology companies, including OpenAI, Anthropic, Google and Meta, reportedly demonstrated the ability to access the internet and penetrate numerous websites.

The incidents have strengthened calls among some security specialists for a more systematic approach to AI risk management. David Robinson, a former OpenAI security specialist, has argued that AI development needs a safety culture comparable to those established in industries such as civil aviation and nuclear energy.

Congress, however, has yet to adopt a comprehensive federal framework. The political dispute has also extended to state-level regulation, with California among the jurisdictions introducing requirements for developers of advanced AI systems, including safety disclosures and reporting of serious incidents.

Industry leaders have attempted to address some of the pressure themselves. On September 29, major American AI companies signed a voluntary code of conduct that includes provisions for independent audits. The agreement does not establish financial penalties, although Trump has described its commitments as morally binding.

The White House has also assigned intelligence chief Jay Clayton a role in examining AI-related threats while seeking to avoid what the administration considers excessive regulation. One alternative under discussion is greater reliance on civil liability, under which companies could face legal consequences if their AI products cause harm.

China: Tight Control Over Applications, Continued Strategic Development

China has taken a different route, combining strong government supervision with an effort to maintain momentum in AI development.

According to analysts at the Oxford China Policy Lab, Beijing's policy does not point toward deliberately slowing the country's technological progress. Instead, authorities have concentrated on controlling how AI is used and the type of content it produces.

AI-generated material is subject to China's broader online information controls and must be identified as AI-generated under rules that took effect in September 2025.

Beijing has also begun imposing specific restrictions on AI-powered virtual companions. Since July, developers have faced rules intended to reduce the risk of emotional dependence, including restrictions involving simulated romantic relationships with minors and requirements encouraging users to take breaks after prolonged interactions.

The rules also prohibit the unauthorized cloning of real individuals.

President Xi Jinping has repeatedly emphasized the need to keep AI under human supervision. In July, he called for legislation, technological monitoring and early-warning and emergency-response mechanisms capable of addressing risks associated with the technology.

At the same time, China's regulatory debate is shaped by geopolitical competition. Analysts note that some Chinese voices see calls for an international slowdown as difficult to reconcile with existing American restrictions on technology exports to China, which have already affected the country's access to advanced AI hardware and technologies.

Chinese authorities have also highlighted the potential use of AI for information manipulation. In September, State Security Minister Chen Yixin accused hostile forces of exploiting AI to produce political rumors.

European Union: A Risk-Based Regulatory Model

The European Union has opted for a more formal legal framework. Its AI Act, adopted in 2024, established a risk-based system that classifies AI applications according to their potential impact and imposes corresponding obligations.

Brussels considers the legislation one of the world's most comprehensive attempts to regulate artificial intelligence. Since August, the European Commission has had the power to impose fines on developers of large AI models, with penalties potentially reaching 3% of their global annual turnover.

Yet implementation has also generated concerns among European businesses. Companies have warned that extensive regulatory requirements could put European developers at a disadvantage against American and Chinese competitors.

As a result, the application of some obligations concerning high-risk AI systems, including technologies used for recruitment and credit decisions, has been postponed until December 2027.

European Commission President Ursula von der Leyen has nevertheless defended the broader strategy, arguing that regulation can give Europe a role in shaping international standards rather than simply reacting to technological developments elsewhere.

Brussels is also examining the risks posed by AI assistants capable of creating highly personalized interactions. In August, ChatGPT became the first AI assistant subject to enhanced obligations under the EU's online-content rules, including annual audits and the possibility of fines reaching 6% of global turnover.

The European Commission has additionally proposed restrictions on chatbots designed to prevent them from creating emotionally dependent relationships with children. The proposal follows concerns surrounding cases in which interactions with AI assistants have been linked to suicides. The measure still requires approval from both the European Parliament and EU member states.

Three Regulatory Philosophies, One Global Challenge

The emerging divide is not simply about whether AI should be regulated. It reflects three different philosophies about where responsibility should lie.

The United States is emphasizing technological leadership, voluntary commitments and liability after potential harm. China is combining centralized supervision of AI applications with an ambition to remain competitive in the global technology race. The European Union is building a legally enforceable framework based largely on the level of risk associated with different uses.

None of these approaches has yet produced a universally accepted model. As AI systems become more capable and increasingly integrated into cybersecurity, employment, finance, communication and personal relationships, governments face the same fundamental challenge: establishing safeguards without undermining the technological development they also consider strategically important.

The next phase of the AI race may therefore be shaped not only by computing power and model performance, but also by the ability of governments to determine where innovation ends and accountability begins.