KRAS drug shifts pancreatic cancer & CAR-T style approach against HIV - News (May 14, 2026)

First, health and medicine—and the headline here is pancreatic cancer. A late-stage study of an experimental drug called daraxonrasib is giving the field something it hasn’t had in a long time: real momentum. The drug targets KRAS, a key driver behind most pancreatic tumors—and many lung and colon cancers too. For decades, KRAS was the poster child for “we can’t hit that with a drug.” Now, with a newer strategy that essentially helps the drug latch on and shut KRAS down, the results are turning heads. In patients with metastatic disease who had already failed chemotherapy, the company reported a median survival of more than 13 months on daraxonrasib, compared with under seven months on standard chemo. It’s not a universal win—responses vary, and side effects can be tough—but if this holds up and gets approved, it could reshape one of the bleakest corners of cancer care.

Staying with breakthroughs, but shifting to infectious disease: researchers are reporting early signs that a modified, CAR-T-like therapy might suppress HIV without daily medication. In a small, early-stage study, two participants who received a single infusion of engineered T cells saw their HIV drop to undetectable levels after stopping standard drugs—lasting nearly a year in one case, and almost two years in the other. This is far from a finished story. The first three participants didn’t respond, and the better outcomes appeared tied to factors like receiving a conditioning chemotherapy dose and starting HIV treatment soon after infection—both linked to a smaller viral “hideout” in the body. Still, with no serious side effects reported so far, it’s a compelling hint that a longer-lasting, potentially “one-and-done” control strategy might be possible—pending larger trials.

And there’s more on HIV—this time on access and manufacturing. Unitaid says it expects to select a South African laboratory to make a generic version of lenacapavir, a long-acting injectable originally developed by Gilead. The big deal here is convenience and discretion: a twice-a-year shot could help people who struggle with daily pills, including those facing stigma or unstable access to care. Unitaid points to evidence that lenacapavir can dramatically reduce transmission risk, and it’s framing the manufacturing plan as part of “medical sovereignty”—building regional capacity so supply doesn’t hinge on distant factories, a vulnerability that became painfully clear during COVID-era vaccine shortages in Africa. If a lab is chosen soon, Unitaid expects production could start within a couple of years—an important timeline to watch.

Next, a major policy signal from Europe: the EU is moving toward tighter online protections for children—and it may include a minimum age for social media, or at least delaying access for younger teens. European Commission President Ursula von der Leyen says the mental-health concerns are too big to shrug off. She’s asked an expert group to deliver recommendations by July, and the Commission could propose new rules as early as this summer. This effort also targets the design of platforms themselves—features like endless scrolling, autoplay, and push notifications that critics say keep kids hooked. The proposed direction would sit under a coming Digital Fairness Act, building on the EU’s existing Digital Services Act. Meanwhile, regulators are already investigating TikTok, X, and Meta over child-safety and age-verification issues. The political pushback is real, including criticism from the U.S. Trump administration—but the EU is signaling it intends to enforce its rules, even against the biggest platforms.

Now to the intersection of AI and global security—where the key theme is speed. A new analysis argues nuclear deterrence depends on a chain of assumptions: no accidents, no uncontrolled escalation, and reliable command-and-control even under pressure. The concern is that AI tools capable of rapidly finding software vulnerabilities could tilt that balance. The piece points to reports about Anthropic’s Claude “Mythos” model allegedly being able to identify and even help exploit previously unknown flaws—so-called zero-days—more quickly. If offensive cyber capabilities get faster and more widely available, critical systems tied to early warning, communications, and decision-making could be at greater risk of disruption or confusion. The fear isn’t a Hollywood-style takeover; it’s misread signals, delayed orders, or degraded information in a crisis—exactly the kind of fog that can turn a tense standoff into a mistake with catastrophic consequences.

In business and technology, chipmaking is getting an even bigger growth forecast. TSMC now predicts the global semiconductor market will surpass one and a half trillion dollars by 2030—up from its earlier one-trillion estimate. The company is pointing to AI and high-performance computing as the main engine of demand, with smartphones and cars still important but no longer the primary growth story. TSMC also says it’s ramping capacity aggressively in 2025 and 2026, including advanced manufacturing and advanced packaging—one of the behind-the-scenes necessities for today’s top AI chips. And it highlighted its expanding global footprint, from the U.S. to Japan and Germany. The takeaway: the AI boom isn’t just about software models—it’s reshaping the physical supply chain of computing, and it’s pulling the entire chip industry upward.

Back to science, with a promising angle on the antibiotic crisis. Researchers at the University of Pennsylvania introduced an AI method called ApexGO designed to speed the search for new antibiotics by repeatedly improving a smaller set of peptide candidates—rather than hunting through massive libraries and hoping for a hit. In lab tests, a large majority of the AI-designed peptides stopped bacterial growth, and many outperformed the originals they were based on. In mouse experiments, two candidates reduced bacterial counts to levels comparable with a last-resort antibiotic. It’s early—questions like safety, durability in the body, and resistance over time remain—but in a world where antimicrobial resistance keeps climbing, a faster, more directed approach to discovery could matter a lot.

A quick stop in human origins—where proteins are stepping in when DNA can’t. Researchers in China analyzed enamel proteins from six Homo erectus teeth using a careful method that preserves the shape of these rare fossils. From samples dating back at least 400,000 years, they found protein variants that may help clarify debated links among Homo erectus, Denisovans, and some modern human populations. One finding is especially intriguing: a variant once thought specific to Denisovans also appeared in these Homo erectus samples, raising the possibility of deeper, more complex mixing among ancient groups than the old family tree diagrams suggest. Even beyond the evolutionary questions, the methods matter—because they offer a way to learn from priceless fossils when DNA is unavailable and sampling must be minimal.

Finally, a cultural and political milestone: the 66th anniversary of the FDA approval of the first oral contraceptive. New coverage looks back at how “the pill” changed American life—giving millions of women far more control over if and when to have children, and reshaping education, careers, and marriage patterns along the way. But it also underscores how contested contraception has been from the start: legal battles, religious opposition, and shifting political winds. And with the post-Roe legal landscape still evolving, some researchers and advocates warn that contraception access could face renewed challenges—even as U.S. fertility remains historically low and the pill is still the most common reversible contraceptive method in the country.