Calls to place tighter controls on the most advanced AI systems have been building across multiple countries, and the three dominant players in that space are each charting a notably different course in response. Policymakers across multiple regions are grappling with a shared set of concerns — among them, AI-driven cyber intrusions, risks to national security, and documented cases in which chatbot interactions preceded the deaths of young users.
The approaches taken by the European Union and the United States diverge from one another. Binding regulations form the cornerstone of the bloc's approach, though lobbying from industry groups led to certain requirements being relaxed before they took full effect. At the federal level in the United States, voluntary commitments and executive directives form the backbone of how artificial intelligence development is governed. China, meanwhile, has been shaping its own regulatory framework, operating under state oversight and setting aside external pressure to moderate the pace of development.
Legislation passed by the European Union in 2024 established a regulatory framework for artificial intelligence that the bloc describes as the first of its kind to address the technology in a thorough and wide-ranging way. Under the legislation, AI systems are sorted into four tiers determined by the degree of risk they present: at one end, low-risk applications such as spam filters and video-game components face few restrictions and are largely free to operate, while at the other end, systems that carry the capacity for broad societal control — including tools used for social scoring, emotional identification, and behavioural manipulation — are prohibited outright.
Industry representatives pushed back against the proposed requirements, arguing the rules would hinder technological development at a time when European firms were already trailing US and Chinese competitors, and ultimately secured a postponement of obligations covering high-risk systems through the AI Omnibus bill, an outcome facilitated partly by a rightward shift in the Commission's direction after the EU elections.
European Commission President Ursula von der Leyen, speaking last month, described the AI regulations as positioning Europe to help determine how such rules take shape at an international level. Plans were also announced to bring together the leading frontier laboratories for a conversation about how their ongoing work at the cutting edge of the field might be better supported. Some firms based in Europe have held back on decelerating, on the grounds that any such move by larger American technology companies could be a tactic to entrench their existing market positions.
An August designation made ChatGPT the first AI-driven chat tool placed on the EU's roster of very large digital platforms, a category carrying heightened regulatory obligations under the bloc's Digital Services Act. A proposal put forward by Brussels last month would require AI companions and chatbots to meet safety standards before being made available to children. Proposed under the EU Kids Act, a prohibition on AI chatbots that foster emotional dependency through simulated personal bonds would take effect if adopted — a measure put forward after a number of deaths by suicide among minors were connected to their use of such tools.
Roughly 20 months into his second administration, President Donald Trump has shaped the country's artificial intelligence agenda primarily through non-binding industry commitments and executive directives aimed at maintaining conditions favourable to AI development and its role in the broader economy. Among Trump's opening actions upon returning to the White House was the cancellation of a 2023 executive order that President Joe Biden had put in place, one that had obligated developers of the most advanced AI systems to hand over safety-testing data to federal authorities and tasked government agencies with drawing up privacy guidelines aimed at curbing algorithmic discrimination.
A plan released by the White House in July 2025 laid out more than 90 policy actions aimed at accelerating AI development, expanding infrastructure investment, and reducing regulatory constraints on the domestic sector. A companion directive carrying the name "Preventing Woke AI in the Federal Government" instructs federal bodies to procure only those artificial intelligence models that have been assessed as free from ideological orientation. The directive rested on no independent assessment framework for neutrality — platforms were simply required to submit their technical records to the relevant authorities.
Congress has yet to produce any artificial intelligence legislation, a gap that reflects both the technology sector's lobbying presence on Capitol Hill and the partisan fault lines dividing members who remain away from Washington until after the November midterm elections. Efforts at the federal level to pre-empt state-level artificial intelligence legislation have not secured congressional backing, leaving statutes such as California's AI Transparency Act to take hold across the country.
Advances in AI that introduced tools capable of conducting sophisticated cyberattacks drew industry attention toward security questions during the year under review. Under an executive order issued in June, technology developers were invited to share their most advanced models with the government on a preview basis — with a window of up to 30 days before public release — under an arrangement that remained entirely optional.
On September 29, leaders from the six largest AI companies in the United States put their names to a White House agreement committing their organisations to external audits and governance oversight at board level for advanced models. Under the arrangement, the firms are left to oversee their own conduct, with no enforcement mechanism attached if those standards go unmet — though Trump described the commitment as one that carries moral weight. A directive issued on that date called on federal agencies to replace the term "artificial intelligence" with "super intelligence" throughout their official written materials.
According to research from the Peterson Institute for International Economics, spending on AI infrastructure and related investment accounted for somewhere between one-fifth and one-third of all US economic expansion in 2026, translating to a contribution of 0.45 to 0.97 percentage points against a broader national growth rate running between 2.1% and 2.3%.
A layered and still-developing body of regulations governs how artificial intelligence may be deployed within China, extending those provisions to cover minors as well. Rules governing AI companion chatbots entered force in July, placing limits on users developing emotional reliance on such systems and prohibiting the construction of digital replicas of identifiable individuals without their knowledge or agreement.
Over the summer, President Xi Jinping put forward a position that artificial intelligence must remain subject to human oversight at all times, alongside a call for expanded legal frameworks, technological monitoring systems, early-warning mechanisms, and structured emergency response protocols. Premier Li Qiang put governments on notice that advances in frontier technologies, AI among them, could outpace the regulatory frameworks meant to govern them if authorities do not act with sufficient speed.
Analysts at the Oxford China Policy Lab noted in September that halting or deliberately slowing down AI advancement does not align with how Chinese policymakers have approached the field. Proponents of this position inside China contend that calls for restraint lack credibility given the trade and technology curbs that have, by their reckoning, constrained the country's AI development over an extended period.
Regulations governing AI-produced content fall under the same framework the country applies to online information more broadly. Chen Yixin, the intelligence head at China's Ministry of State Security, stated the previous month that AI tools posed a threat to national security, adding that adversarial actors were exploiting generative AI technologies to manufacture politically charged rumours and circulate misleading content.
