Governments and Vatican move to shape frontier AI
White House plans pre-release government review of frontier AI models; EU opens high-risk classification consultation; Vatican prepares AI ethics document.
By BINA Editorial
From Washington to Brussels and Rome, this week brought a convergence of official efforts to set boundaries around the most powerful AI systems yet built.
White House prepares executive order on frontier AI oversight
The White House is expected to sign an executive order requiring AI developers to share advanced models with federal agencies before their public release. Under the proposed framework, companies would provide access to frontier models 90 days before launch, with the National Security Agency assigned to conduct classified testing for cybersecurity risks. The initiative marks a notable shift for the Trump administration, which previously favoured a hands-off approach to AI policy; some industry players are pushing for a shorter 14-day window. The order also includes provisions to strengthen cybersecurity across government networks and expand the US Tech Force. (Politico | Axios | Reuters)
EU Commission opens public consultation on high-risk AI classification
On 19 May, the European Commission published draft guidelines to help businesses and public bodies determine whether their AI systems fall into the "high-risk" category under the EU AI Act. The document — delayed from its original February deadline — covers areas including biometrics, education, employment, and law enforcement, with a public consultation open until 23 June 2026. The guidelines accompany the May Digital Omnibus agreement, which pushed compliance deadlines for high-risk systems back to December 2027 and August 2028, giving operators more time to adapt. (JD Supra | White & Case | DataGuidance)
Vatican prepares landmark AI ethics encyclical
Pope Leo XIV will personally present his first major teaching document on artificial intelligence — titled Magnifica Humanitas ("Magnificent Humanity") — on 25 May. He will be joined by Christopher Olah, co-founder of Anthropic. The encyclical focuses on "the protection of the human person in the age of artificial intelligence," and signals the Vatican's intention to serve as a sustained moral voice in debates about AI development, safety, and governance — particularly on the rights of those most vulnerable to automated systems. (National Catholic Reporter)
NHS launches £900m AI procurement framework; Imperial College develops infection-prediction AI
NHS Shared Business Services launched a £900 million "Healthcare AI Solutions" framework on 11 May, covering AI procurement for NHS trusts and wider public-sector bodies from May 2027 through 2035. It spans diagnostics, predictive analytics, surgical robotics, and operational tools — with the stated goal of making the NHS the first national health system to routinely deploy AI. Separately, researchers at Imperial College London announced an AI platform — developed with €1 million in pre-seed funding — that can forecast highly resistant hospital infections days in advance by modelling transmission pathways within wards.
Nvidia posts $58bn quarter on AI chip demand; Vera Rubin system due H2 2026
Nvidia reported $58.32 billion in revenue for its fiscal first quarter of 2026, up from $18.78 billion in the same period a year earlier — far exceeding Wall Street expectations. Demand for AI accelerator chips drove virtually all of the growth. CEO Jensen Huang said the company expects to be "supply-constrained throughout the entire life" of its forthcoming Vera Rubin AI system, described as a generational leap in infrastructure and scheduled for the second half of 2026. (Washington Post | The Guardian)
Global study finds widening gap between employer AI plans and worker readiness
A global workforce study by Adecco Group, released today, found that 45% of business leaders plan to integrate AI agents into workflows within the next 12 months, yet only 36% believe their organisation's talent strategy clearly demonstrates that AI will create opportunities for employees. Separate data from Staffing Industry Analysts, published 18 May, shows that 17 US occupations highly exposed to AI automation saw combined employment fall 1.6% between May 2024 and May 2025 — with customer-service roles down 4.8% and secretarial roles (excluding medical and legal) down 1.8%. (PR Newswire | Staffing Industry Analysts)