AI weekly roundup, 16-17 May 2026

Sunday, 17 May 2026

Anthropic warns: China could overtake the US in the AI race by 2028

Anthropic released a policy paper this week laying out two possible futures for 2028, the year by which the company expects transformative AI systems to arrive. In the first scenario, the US and its democratic allies preserve a 12-to-24-month lead in frontier model development by tightening export controls on advanced chips, deflecting "distillation attacks" from Chinese firms, and accelerating American AI adoption in global markets. In the second, without decisive US action, Chinese firms close the gap and global AI norms are shaped by authoritarian regimes. The team's analysis projects Huawei producing only about 4% of Nvidia's aggregate output in 2026, falling below 2% in 2027.

Pope Leo XIV signs his first encyclical, dedicated to artificial intelligence

Last Friday, Pope Leo XIV signed his first encyclical — exactly 135 years after Pope Leo XIII signed Rerum Novarum, the encyclical on workers' rights during the Industrial Revolution that became a cornerstone of modern Catholic social teaching. The new encyclical is expected to position artificial intelligence within the Church's social teaching across labour, justice, and peace. Public release is expected in the coming weeks and is anticipated to become another point of friction between the Holy See and the Trump administration, which has rolled back bureaucratic hurdles to rapid AI development.

Google I/O 2026 opens in two days — anticipation builds across tech

Google opens its annual developer conference on Tuesday, 19 May, at the Shoreline Amphitheatre in Mountain View, California. The keynote will cover sweeping updates to the Gemini model family, deep details on Aluminium OS — the Android-derived operating system for the new Googlebook laptops — innovations in Android XR, and a major push for agentic AI. Last week Google previewed "Gemini Intelligence" as a new intelligence layer in Android and announced the Googlebook category in partnership with Acer, ASUS, Dell, and Lenovo. Tuesday's headline announcement is reserved for the new Gemini model.

Google updates its SEO guidance for the AI Overviews era

Google's Search Central published an updated guide this week for writers and marketers working through the consequences of AI Overviews. The new guidance clarifies that the fundamentals of search optimisation remain relevant in the AI era, and that Google's AI system relies on RAG and a technique called Query Fan-Out, both backed by the standard ranking pipeline. Google recommends avoiding dedicated llms.txt files, excessive chunking, and overloaded structured data, and instead focusing on original content with a distinctive angle that AI can't easily reproduce. A Pew study found click-through on result pages with AI Overviews dropped to just 8%, down from 15% on pages without the summary; an Ahrefs study found click-through reductions of up to 58%.

Marc Benioff: Salesforce will spend $300M on Anthropic this year

Salesforce CEO and founder Marc Benioff disclosed on the All-In podcast this week that the company expects to spend roughly $300 million this year on Anthropic tokens, primarily for code and product development. "These code agents are excellent. Anthropic is excellent," he said. The statement joins a string of confirmations from major enterprises adopting Claude Code as their internal standard — a pattern reinforced by the new AI Ramp Index data, which shows 44.4% of measured US companies using Anthropic versus 32.3% on OpenAI.

Saturday, 16 May 2026

Pope Leo XIV establishes a dedicated AI research group at the Vatican

The Vatican announced on Saturday the establishment of an internal research group under Pope Leo XIV focused on artificial intelligence, a day after the Pope signed his first encyclical. The announcement explained the decision was taken "in light of the acceleration in AI use, its potential consequences for individual persons and humanity as a whole, and the Church's concern for the dignity of every person." The Pope, a Chicago native with a degree in mathematics, has previously warned about AI use in weapons in the Middle East and Ukraine, and about generative AI's capacity to deceive through deepfakes. He has also warned priests against using AI to write sermons.

Malta — first country in the world to give every citizen free ChatGPT Plus

OpenAI and the Maltese government announced on Saturday an unprecedented partnership: every resident of the country will receive a free year of ChatGPT Plus after completing an AI literacy course. The course, developed by the University of Malta under the name AI for All, covers the basics of the technology, its capabilities and limits, and responsible use at home and at work. Eligibility verification runs through the country's online identity system. The program opens this month and will expand as more residents complete the course. Maltese citizens living abroad are also eligible. Economy Minister Silvio Schembri stated that Malta is the first because the country refuses to leave its citizens behind in the digital era. This is the first time OpenAI has signed a partnership at the national level with an entire country.

Anthropic calls for sharp tightening of export controls on China

In its new policy paper, Anthropic urges Washington to close loopholes in export controls on advanced AI chips, restrict remote access to controlled compute infrastructure, and block "distillation attacks" in which Chinese firms copy capabilities from frontier models through structured queries. The company stated that its support is for the Chinese people and the Chinese research community, not for the rule of the Communist Party. In January 2026, the US House of Representatives passed a bipartisan bill 369-22 designed to close this loophole, but the legislation has not yet been ratified by the Senate.

TIME analysis: AI was the elephant in the room at the Trump-Xi summit

TIME magazine published an analysis on Saturday arguing that despite the impressive technological composition of President Trump's delegation to China — Elon Musk, Tim Cook, Jensen Huang, and Larry Fink — the two presidents focused their talks on trade and reached no agreements on the future of artificial intelligence. The analysis suggests Trump missed a rare opportunity to shape the economic and strategic terms of the AI race between the two powers. The summit closed Friday without formal agreements, but the backdrop — approval for export of H200 chips to ten Chinese companies — has already affected markets. Democrats in Congress have expressed concern over the sale's approval.

Anthropic overtook OpenAI in the US enterprise market per the AI Ramp Index

Publication of the AI Ramp Index for Q1 2026 showed Anthropic crossing OpenAI for the first time as the leading AI provider among American companies. 44.4% of measured companies use Anthropic's tools versus 32.3% on OpenAI. Anthropic has quadrupled its presence within a year. Fortune's analysis explains the success stems from Anthropic's focus on enterprise tools rather than a consumer chatbot — primarily Claude Code for autonomous development and the Mythos model for software vulnerability detection.

The Vatican: AI must complement human intelligence, not replace it

Alongside the announcement of the new research group, the Vatican emphasised a message that has run through Catholic engagement with AI since the Church entered the public debate — the technology must be a tool that complements, not replaces, human intelligence. The Vatican also warned of "the vast quantities of energy and water" consumed by AI data centres, and called for examination of the environmental impact of the race. In 2020, Microsoft, IBM, and Cisco signed the Rome Call for AI Ethics, and Pope Francis in his later years called for an international convention to regulate AI.