In a week that saw seismic shifts across the AI landscape — from the halls of the Vatican to the corridors of the White House — the artificial intelligence industry finds itself at a crossroads between ambition and accountability. Here are the biggest stories shaping the field.
Vatican Releases Historic AI Encyclical With Anthropic's Christopher Olah
Pope Leo XIV today presents Magnifica Humanitas ("Magnificent Humanity"), his first encyclical and the Catholic Church's most comprehensive statement on artificial intelligence to date. The document, which centers on "the protection of the human person in the age of artificial intelligence," is being unveiled at the Vatican's Synod Hall alongside a notable guest: Christopher Olah, co-founder of Anthropic and a leading researcher in AI interpretability.
The pairing is striking. Olah's work on mechanistic interpretability — understanding what happens inside neural networks at a granular level — aligns with the encyclical's core concern: ensuring humans can understand, govern, and remain central to the AI systems reshaping society. The Vatican's choice to present the document alongside a figure from the AI safety community, rather than a tech CEO, signals a deliberate emphasis on transparency and human dignity over commercial interests.
The encyclical arrives at a particularly charged moment for Anthropic, which has been labeled a national security concern by the Trump administration after refusing to grant the Pentagon unrestricted use of its AI models for lethal autonomous weapons and mass surveillance.
Trump Abruptly Scraps AI Executive Order Hours Before Signing
In a dramatic policy reversal, President Trump on May 21 canceled the signing of what would have been his administration's first major AI regulation executive order — just hours before the ceremony was set to begin, with tech executives already in Washington.
The order would have created a framework for the federal government to access and evaluate unreleased frontier AI models, with a provision requiring labs to share models 90 days before public release. It was itself a remarkable about-face: the same administration had revoked the Biden-era AI safety executive order upon taking office, arguing it would slow American innovation.
The catalyst for even considering the order was Anthropic's Mythos model, which demonstrated the ability to autonomously discover thousands of severe software vulnerabilities, including flaws in major operating systems and web browsers. But Trump ultimately pulled back, citing concerns about ceding the U.S. lead over China. The episode underscores the administration's internal tension between its deregulatory instincts and the growing national security implications of frontier AI capabilities.
Andrej Karpathy Joins Anthropic in Blockbuster Talent Move
In one of the most significant talent moves in recent AI history, Andrej Karpathy — OpenAI co-founder, former Tesla AI chief, and one of the most respected researchers in the field — announced he has joined Anthropic.
Karpathy is working on pre-training under team lead Nick Joseph, and is starting a new team focused on using Claude to accelerate pre-training research itself — a recursive approach that uses the AI to improve the process of building AI. "I think the next few years at the frontier of LLMs will be especially formative," Karpathy wrote on X. "I am very excited to join the team here and get back to R&D."
The move is notable not just for who Karpathy is, but for where he came from. As one of OpenAI's founding members, his decision to join its chief rival speaks volumes about Anthropic's growing gravitational pull in the AI talent war. After leaving OpenAI in 2024 to start Eureka Labs, an AI-powered education startup, Karpathy's return to frontier research suggests that the current moment in AI development is too consequential to sit out.
Google Overhauls Search With AI-First Redesign
At Google I/O 2026 on May 19, Google unveiled what it called the biggest update to Search in 25 years: a complete AI-first redesign powered by Gemini 3.5 Flash. The search box itself has been reimagined — it now expands dynamically, accepts multimodal inputs including images, PDFs, and video, and offers AI-powered suggestions that go far beyond traditional autocomplete.
The upgrade brings AI Mode, previously a limited feature, to all users globally. Powered by Gemini 3.5 Flash — which Google says delivers "Pro-level" reasoning at Flash-class speed — AI Mode transforms Search into a conversational experience where users can ask detailed questions and receive synthesized, contextual answers.
Perhaps most significantly, Google introduced Search agents: customizable AI agents that operate in the background to find and track information on behalf of users. It's a bold bet that the future of search isn't typing queries, but delegating tasks — and it marks Google's most aggressive move yet to defend its core business against the rising tide of AI-native competitors.
Google also cut the price of its AI Ultra subscription from $250/month to $200/month and introduced a new $99/month tier, both including access to the full Gemini model stack and the new Gemini Spark personal AI agent.