US federal and state agencies are signalling a potential shift towards a more proactive approach to Artificial Intelligence regulation, particularly with regard to security. Earlier this year, President Donald Trumpʼs administration sought to override state-level Artificial Intelligence regulations in pursuit of an ultra-light touch national standard. However, the US Department of Commerceʼs Center for Artificial Intelligence Standards and Innovation is now conducting pre-deployment evaluations and research to assess the state of security at Google Deepmind, Microsoft and xAI.
The center, formerly the US Artificial Intelligence Safety Institute and now part of the National Institute of Standards and Technology, acts as the main liaison point between government and the technology industry. To date, CAISI has completed more than 40 Artificial Intelligence evaluations, according to its May 5 statement. The agreement with the three providers is intended to support voluntary improvements and strengthen international Artificial Intelligence competition, while also sharpening understanding of frontier systems and their national security implications.
Corlytics data suggests US regulators have been discussing the role of Artificial Intelligence in guidance and consultation papers for years, but without consistently moving toward formal rulemaking. Since the beginning of 2015, 15 federal-level documents have mentioned Artificial Intelligence, while a further 22 state-level documents have addressed the technology. Some of those documents identified risks linked to Artificial Intelligence, but did not point to a clear regulatory precursor.
Internationally, the pattern is different. The UK Financial Conduct Authority has mentioned Artificial Intelligence in only nine of its regulatory documents since January 2025, reflecting its no Artificial Intelligence regulation stance and preference for sandboxes. In Europe, the EU Artificial Intelligence Act applies across the bloc, with uses in critical areas including education, recruitment, some elements of financial services and law enforcement classified as high-risk under Annex III and subject to mandatory assessments under Article 43.
Other jurisdictions are leaning on sector-specific measures. Singapore has been issuing guidance through industry regulators such as the Monetary Authority of Singapore and set out Artificial Intelligence missions in its 2026 budget statement for sectors including finance, manufacturing and healthcare. Japan passed the Artificial Intelligence Promotion Act last year to support research and development while using sector-specific guidelines in high-risk areas. A Bank for International Settlement working paper also found the most highly valued Artificial Intelligence firms are concentrated in the US and China, followed by the euro area, Taiwan, South Korea, Japan, India, the UK, Canada and Australia.
