Some colleagues read Leo XIV's first encyclical as a policy document and came away disappointed. I think it's wrestling with something deeper: the dominant AI narrative says capability is the answer, but Augustine would ask whether being known, being loved, and having purpose were ever capability problems in the first place.

The main theme at this year’s Google I/O is that the next phase of AI isn't about the extraordinary, it's about the everyday.

Most of the conversation so far has been about frontier capabilities: models that pass the bar, write production code, do PhD-level research. Google spent the week pointing somewhere else. The through-line across 100+ announcements wasn't a better benchmark score — it was the parent on snack duty, the inbox that manages itself, the search that finishes the task instead of just handing back links. Agents built not for developers shipping code faster, but for the rest of us who just want life to feel a little simpler.

The most profound way I used AI in 2025 wasn't at work. It was during one of the hardest stretches my family has faced, when my mother's cancer came back. I built an assistant that could translate pathology reports into plain language, reason through her case like a tumor board, and carry her full medical history into rooms where the EHRs couldn't talk to each other. And I built something else: a chatbot she could ask anything, at any hour, without feeling like a burden. This isn't a story about AI curing cancer — cancer is navigated decision by decision. But in a fight measured in small gains, clarity was its own kind of gift.

As large language models (LLMs) increasingly replace traditional search engines as tools for information gathering, the use of AI in the political arena—and its impact on elections—is inevitable. Recent research published in Nature and Science suggests that AI chatbots are not merely passive sources of election information; they can actively shape voter attitudes in measurable and durable ways.

I was honored to testify before the Senate HELP Committee, in my capacity as a Non-Resident Senior Fellow at AEI, on how AI can strengthen support for patients, workers, children, and families. It is no secret that I am an optimist about AI’s promise to expand opportunity and make learning more personal. But optimism must be matched with responsibility, including being honest about the emerging challenges and risks. As AI becomes more persuasive and more present in classrooms, we need to take seriously the risks of over-reliance, misalignment with educational goals, and emotional dependence on systems designed to imitate empathy.

A recent webinar with Julia Freeland Fisher, Ryan McBain, and Pat Pataranutaporn explored how AI is learning to simulate warmth, patience, and compassion—traits that make systems feel empathetic. These advances promise major benefits, including personalized tutoring, mental-health support, and reduced loneliness, but also raise concerns about dependency, weakened human bonds, and manipulative “addictive intelligence.” Experts urged designing AI that promotes human flourishing, measuring emotional and social well-being—not just accuracy—and enforcing transparency and child-safety standards. The central question: can AI empathy strengthen human connection rather than replace it?

OpenAI’s GPT-5 introduces a breakthrough in AI safety: real-time monitoring of its internal reasoning, or chain-of-thought (CoT). GPT-5 cut deceptive reasoning from 4.8% in earlier models to 2.1%, with 81% precision and 84% recall in flagging risks. Yet studies warn this interpretability may erode as models evolve toward opaque, machine-optimized reasoning. Both OpenAI and academic researchers urge treating CoT readability as a core safety metric—measured, reported, and preserved through training—to ensure humans retain visibility into how advanced AI systems reason, decide, and potentially deceive.

The White House’s 2025 AI Action Plan outlines an ambitious roadmap aimed at securing America’s global leadership in artificial intelligence through rapid innovation, strategic infrastructure, and assertive international engagement. Emphasizing open-source models, regulatory sandboxes, and workforce initiatives, the plan boldly contrasts previous approaches by accelerating deployment and competitiveness over precautionary regulation. However, critical gaps remain in AI interpretability, state-federal dynamics, copyright clarity, and execution specifics. The ultimate measure of success will depend not merely on technological advancement, but on ensuring responsible governance, maintaining public trust, and effectively managing AI’s profound risks alongside its transformative opportunities.

Google I/O 2025 underscored a pivotal shift toward the "Intelligence of Things," embedding advanced AI into everyday tools and devices, seamlessly integrating technology into daily life rather than creating isolated products. Google's vision emphasizes proactive, personalized AI, exemplified by Gemini 2.5’s deep research tools and immersive communication innovations like Beam and real-time multilingual translation in Google Meet. The event highlighted AI's transformative impact on education, notably through Project Astra's multimodal learning capabilities and LearnLM’s integration into Gemini, significantly enhancing pedagogical effectiveness. Central to this evolution is the critical role of agile public policy in facilitating safe innovation, ensuring AI technologies serve human needs by enhancing safety, productivity, and quality of life.

Meta was kind enough to extend an invitation for me to attend its inaugural LlamaCon—a one-day developer summit devoted to the Llama family of open-source large language models. It offered the chance to better understand the direction in which both the technology and its surrounding ecosystem are moving, and therefore merits a close read by anyone shaping AI strategy or policy.

AI is advancing faster than anyone expected — and the 2025 AI Index Report proves it. From AI models matching Nobel-level intellect to a 280x drop in computing costs, Stanford’s latest report maps the jaw-dropping acceleration of AI capabilities. The U.S. still leads in AI models and private investment, but China is closing the quality gap—and fast. Meanwhile, AI-driven labs are already making major scientific breakthroughs, like designing nanobodies to fight COVID-19, and outperforming doctors in diagnostic accuracy.

AI is transforming the workplace by enhancing collaboration, speeding up tasks, and expanding access to expertise. A Harvard-Wharton-P&G study found individuals using AI performed as well as human teams, and AI-assisted teams were nearly three times more effective. AI also helped less experienced employees contribute more and encouraged interdisciplinary thinking. Pennsylvania’s ChatGPT pilot showed similar benefits, with employees saving 95 minutes daily and improving workflows. A national survey by JFF found that 57% of workers already see AI reshaping their jobs, though few receive formal training. These findings suggest AI isn’t just a tool for efficiency—it’s a partner reshaping how teams work, learn, and innovate.

In 2025, your next coworker might not be human, but an autonomous AI agent capable of performing specialized, real-world tasks collaboratively and independently. Groundbreaking experiments from Google, Stanford, and Chan Zuckerberg BioHub reveal that teams of AI agents—ranging from computational biologists to critical reviewers—can rapidly innovate, discover new medical treatments, and respond dynamically to emerging. As AI becomes embedded into workplace dynamics, businesses face an exciting yet challenging future, where managing AI agents as digital employees might soon become as common as managing human teams.

DeepSeek's R1 model from China has sent shockwaves through the AI industry, matching OpenAI's top models while claiming development costs of just $5 million. Though experts dispute this figure, R1's emergence has triggered market value losses and security concerns. US companies quickly responded, with Google updating Gemini to outperform R1. The situation highlights tensions around export controls, security vulnerabilities, and the strategic importance of open-source AI development that aligns with democratic values. Security risks have led several US entities to ban DeepSeek. America must accelerate innovation and education to maintain AI leadership.

AI has become an invaluable partner for me, giving me instant access to a wide array of expert “assistants.” I now have a data analyst, driver with Waymo, brainstorming partner, legislative analyst, medical assistant, start-up advisor, graphics designer, and researcher at my fingertips, ready to help whenever I need specialized skills and knowledge.

A recent social media clash that erupted between Elon Musk, Vivek Ramaswamy and Trump loyalists over high-skilled immigration reform exposed deep ideological rifts within the Republican coalition. But the importance of the debate over immigration policy and the American education system extends far beyond social media — solving these problems is critical to America’s competitiveness. By combining pragmatic immigration reforms, bold educational investments and innovative AI-driven learning, we can forge the “Talent Dominance” agenda we desperately need. 

In an analysis of Sal Khan's "Brave New Words" and the evolving landscape of AI in education, I present a case for AI-powered tutoring as a transformative force. Recent advancements in speech, image analysis, and emotional intelligence, combined with promising research studies showing significant learning gains, suggest AI tutoring could help address our urgent educational challenges like pandemic learning loss and chronic absenteeism.

I was excited to contribute to a group of over 20 prominent AI researchers, legal experts, and tech industry leaders from institutions including OpenAI, the Partnership on AI, Microsoft, the University of Oxford, a16z crypto, and MIT on a paper proposing "personhood credentials" (PHCs) as a potential solution to the growing challenge of AI-powered online deception. PHCs would provide a privacy-preserving way for individuals to verify their humanity online without disclosing personal information. While implementation details remain to be determined, the core concept warrants serious consideration from policymakers and tech leaders as AI capabilities rapidly advance, threatening to erode the trust and accountability essential for societies to function.

The recent AI policy roadmap from the bipartisan AI working group in Congress strikes a thoughtful balance between promoting AI innovation and addressing potential risks. It lays out a nuanced approach including increased AI safety efforts, crucial investments in domestic STEM talent, protections for children in the age of AI, and "grand challenge" programs to spur breakthroughs - all while avoiding hasty overregulation that could stifle progress