OB-GYN Group Breaks From CDC On Pregnancy Vaccines: The American College of Obstetricians and Gynecologists has issued its own vaccine schedule for pregnant people, marking its first immunization guidance independent of the CDC. ACOG recommends COVID-19, RSV, influenza, and Tdap vaccines during pregnancy, continuing its previous evidence-based position even as CDC guidance has shifted. The split reflects growing tension between professional medical societies and federal vaccine policy, especially after the CDC dropped its COVID-19 vaccine recommendation for pregnant women and removed ACOG experts from advisory workgroups. ACOG says the new schedule is meant to reduce confusion and preserve science-based care. Thirteen medical organizations, including the American Academy of Pediatrics and Infectious Diseases Society of America, endorsed the schedule, emphasizing maternal vaccination’s role in protecting newborns. (CIDRAP)

China’s JUNO Takes Lead In Neutrino Race: China’s Jiangmen Underground Neutrino Observatory has delivered an early signal that it may solve one of particle physics’ biggest puzzles: the ordering of neutrino masses. Reporting in Nature after just 59 days of data, JUNO measured two neutrino oscillation parameters with unmatched precision, cutting their uncertainty by about one-third. The $300 million detector watches electron antineutrinos from nuclear reactors 53 kilometers away, using 20,000 tons of flashing liquid and more than 43,000 phototubes to map their energy spectrum. Subtle ripples in that spectrum could reveal whether neutrinos have two light mass states and one heavy one, or the reverse. The answer could shape future studies of supernovas, cosmic evolution, and whether neutrinos are their own antiparticles. (Science)

Trump Floats Public Stake In AI Boom: President Trump said he expects major artificial intelligence companies to “give back” to the public, suggesting the government could seek some form of stake in leading AI firms. Speaking at the White House, Trump said he planned to meet with 12 to 15 top AI executives soon. The idea appears tied to broader anxiety over AI’s economic impact: Reuters noted that a recent Reuters/Ipsos poll found half of Americans fear AI could cost them or someone in their household a job. OpenAI has separately proposed a public wealth fund tied to AI profits, and Sen. Bernie Sanders has expressed support for similar thinking. The story signals that AI policy may be moving from regulation toward redistribution. (Reuters)

AI Debt Boom Could Hit $570 Billion: Morgan Stanley expects global AI-related debt issuance to more than double in 2026, reaching nearly $570 billion as hyperscalers seek new ways to finance enormous infrastructure spending. Reuters reports that AI-related debt issuance had already reached nearly $236 billion by May 31, roughly four times the level from the same period last year. Alphabet, Amazon, Microsoft, and Meta are expected to spend about $700 billion this year, with capital expenditures potentially passing $1 trillion in 2027. The shift suggests that the AI buildout is no longer funded only by cash-rich technology balance sheets. It is becoming a credit-market story, too, with bond investors increasingly exposed to the economics of data centers, chips, and compute demand. (Reuters)

Google Turns World Cup Into AI Testbed: Google’s Gemini will quietly enter the 2026 World Cup through a partnership with Argentina’s national team, giving players and coaches access to AI tools for tactical analysis. WIRED reports that the models will be used to break down plays, analyze opponents, and potentially shorten the time between scouting and action. Google has not fully detailed the internal systems Argentina will use, but the stakes are clear: elite soccer offers a live, high-pressure environment for testing whether AI can produce practical competitive advantages. The story is less about consumer chatbots than applied decision support, where milliseconds, formation changes, and opponent tendencies matter. If successful, this could accelerate AI adoption across professional sports, coaching staffs, and performance departments. (WIRED)

Apple Reboots Siri Around AI: Apple used WWDC 2026 to reintroduce Siri as an AI-heavy assistant, now called Siri AI, after delays and criticism over earlier Apple Intelligence promises. The Verge reports that Siri AI will support more detailed conversations, onscreen awareness, image editing, writing tools, and conversation history through a standalone app. Apple also showed AI updates across iOS 27, macOS 27, Safari, Messages, Mail, Phone, Photos, and the Home app. Safari will use AI to organize tabs and generate extensions from natural-language prompts, while macOS will put Siri directly into Spotlight. The move matters because Apple is embedding AI at the operating-system level rather than treating it as a separate app, potentially reshaping how hundreds of millions of users encounter generative tools. (The Verge)

Google Rents SpaceX Compute Capacity: Google will pay SpaceX $920 million per month from October 2026 through June 2029 for access to roughly 110,000 Nvidia GPUs, CPUs, memory, and related infrastructure, according to TechCrunch. The deal, disclosed in a SpaceX filing, follows a similar agreement in which Anthropic committed $1.25 billion per month for Colossus 1 compute capacity. Google described the arrangement as short-term “bridge capacity” to meet stronger-than-expected demand for Gemini Enterprise and its agent platform. The deal is striking because Google is already one of the world’s largest AI compute owners, yet still needs outside capacity. It also illustrates how AI infrastructure is becoming a strategic asset in its own right, linking cloud platforms, model developers, and space-tech companies. (TechCrunch)

Anthropic Expands Through TCS Partnership: Anthropic has partnered with Tata Consultancy Services to scale Claude deployments across enterprise customers, according to TechCrunch. TCS will create a dedicated business unit focused on Anthropic’s AI models, gain early access to new releases, and provide Claude to more than 50,000 employees. The companies plan sector-specific solutions for finance, healthcare, telecommunications, and aviation. TCS businesses will also use Claude for customer service, process automation, training, certification, claims adjudication, and lending advisory tools. The deal reflects a broader race among frontier AI companies to secure distribution through major IT services firms, especially in India, which Anthropic has described as its second-largest market. It also shows how AI is pressuring traditional outsourcing firms while simultaneously giving them new products to sell. (TechCrunch)

AI-Designed Coronavirus Vaccine Clears Human Trial: An AI-designed universal coronavirus vaccine has passed its first human clinical trial, according to ScienceDaily’s report on University of Cambridge work. The vaccine, developed by Cambridge researchers and DIOSynVax, was tested in 39 healthy volunteers and found to be safe, well tolerated, and free of significant side effects. Unlike strain-specific vaccines, it targets shared features across the Sarbeco coronavirus family, including SARS-CoV-2, SARS, and related bat viruses with pandemic potential. The trial found immune responses against multiple coronaviruses, including viruses that have not yet infected humans. The study, published in the Journal of Infection, is also notable because the vaccine’s active ingredient was created entirely through computer simulations, making it a concrete example of AI-assisted pandemic preparedness. (ScienceDaily)

AI May Speed New Physics, But Miss Surprises: AI could help cosmologists search for new physics faster, but it may also become too anchored to what it already knows. ScienceDaily reports on research showing that transfer learning can reduce the need for costly simulations when studying cosmic structure and possible departures from standard models. The catch is that a model trained on familiar patterns may underperform when the universe behaves in a genuinely unfamiliar way. In other words, the same shortcut that makes AI efficient can make it less sensitive to anomalies. The finding is important for AI-driven science because discovery often depends on noticing the unexpected. The work suggests future scientific AI systems may need mechanisms for “forgetting,” uncertainty, or deliberately questioning inherited assumptions. (ScienceDaily)

Sepsis AI Studies Hid A Timing Flaw: Emory University researchers found a serious flaw in many reinforcement-learning studies aimed at guiding sepsis treatment. EurekAlert reports that a common preprocessing and indexing method can create a slight time misalignment, allowing an AI agent to use future events to predict the past. If test data contain the same flaw, performance metrics can look artificially strong while hiding real-world failure. The team warned that flawed systems could recommend overtreatment or undertreatment in nearly half of patient states if deployed clinically. The finding is especially sobering because Tang and colleagues said the mistake appeared across much of the literature, including their own earlier work. The group also proposed a workaround, making this both a warning and a corrective step for medical AI. (EurekAlert!)

Brain-Inspired AI Uses Less Energy: Researchers at the University of Massachusetts Amherst have developed a brain-inspired AI approach designed to learn continuously while using far less computing energy than conventional systems. EurekAlert reports that the work, led by AI pioneer Hava Siegelmann and published in Nature Communications, aims to move beyond fixed training phases and toward systems that adapt in real time. That matters because today’s large AI systems are increasingly constrained by data-center energy use, hardware costs, and the need for repeated retraining. The new approach draws on aspects of biological computation, where brains learn continuously without consuming anything like the power demanded by frontier models. It is still research, not a commercial platform, but it points toward AI that may be more adaptive, efficient, and sustainable. (EurekAlert!)