Blogs

Inside your intestinal lining sits an arsenal of microscopic syringes. They belong to bacteria that pose no threat—at least not intentionally. But they’ve been actively sabotaging your immune responses, shaping inflammation, and potentially steering you toward disease. And we only just noticed what they’ve been doing. For decades, we’ve known that changes in gut bacteria correlate with everything from inflammatory bowel disease to allergies to obesity. But knowing something changes and understanding why it changes are entirely different things. “Our goal was to better characterise some of the underlying processes of how gut bacteria affect human biology,” says Veronika Young, a researcher at Helmholtz Munich who led the investigation. “By systematically mapping direct protein–protein interactions between bacterial and human cells, we can now suggest molecular mechanisms behind these associations.” The scale of what they found is still sinking in. Nearly 80 per cent of the common Gram-negative bacteria living peacefully in healthy guts possess these injection systems—the kind of molecular machinery that scientists have long assumed belonged exclusively to dangerous pathogens like Salmonella and Yersinia. Until now, nobody knew the harmless residents had this arsenal. These aren’t elaborate systems. Under a microscope, they look deceptively simple: microscopic needle-like structures that bacteria use to pierce cell membranes and squirt proteins directly into host cells. It’s elegant evolution, really. The bacteria essentially weaponise themselves at the molecular level, delivering customised chemical commands into the very heart of our cellular machinery. When Pascal Falter-Braun, director of the Institute for Network Biology at Helmholtz Munich, first started examining these systems in benign bacteria, his immediate reaction was striking: “This fundamentally changes our view of commensal bacteria. It shows that these non-pathogenic bacteria are not just passive residents but can actively manipulate human cells by injecting their proteins into our cells.” This isn’t subtle sabotage, either. The bacteria have targets, and those targets are striking. When researchers created a massive map of interactions—over a thousand distinct connections between bacterial proteins and human immune molecules—a pattern emerged with unsettling clarity. The bacterial proteins weren’t randomly probing the human cell. They were zeroing in on the master control switches of immunity itself. Chief among these targets: the NF-κB signalling pathway, one of the oldest immune circuits in the animal kingdom, and the molecules that control cytokine production—the chemical messengers that coordinate your entire immune response. “The bacterial proteins preferentially target human pathways involved in immune regulation,” Young explains. In other words, the bacteria have learned exactly which molecular buttons to push to shape how your immune system behaves. Here’s where the story gets genuinely unsettling. The team discovered that these effector proteins—the cargo the bacteria inject—are enriched in the microbiomes of people with Crohn’s disease, one of the most debilitating forms of inflammatory bowel disease. Not just present. Enriched. The bacteria weren’t neutral passengers accumulating by chance. They were driving something. They tested this hypothesis directly. When they introduced these bacterial proteins into human cells growing in the laboratory, something happened: the cells shifted their inflammatory response. The bacteria had literally rewired how the human immune system signalled, reducing some inflammatory molecules whilst ramping up others depending on the molecular context. It wasn’t random. It was controlled manipulation. The findings are making researchers rethink nearly everything about how we conceive of commensal bacteria. For more than a century, microbiology operated under a simple mental model: there are pathogens (bad) and commensals (neutral hangers-on). But evolution rarely works in such binaries. The bacteria that evolved these injection systems didn’t do so yesterday. These systems have been refined across millions of years. Falter-Braun and his colleagues found evidence that the effectors in benign gut bacteria follow entirely separate evolutionary trajectories from their pathogenic cousins. The structures were distinct. The sequences were distinct. The domains they possessed were unique to commensals.   Discover more Health and wellness retreats Science news digest Science museum memberships Blogging platform access Space exploration news Wildlife conservation donations Educational science kits Health & Non-humans Mathematics tutoring services Science communication workshops   “Many of these effectors appear designed to support a non-pathogenic lifestyle,” Young notes. Some of them encode proteins that mess with a bacterial signalling molecule called cyclic diguanylate—the kind of system you’d expect bacteria to use internally to sense their environment and coordinate behaviour, not to attack hosts. Others seemed calibrated to nudge human immune responses in ways that might actually benefit bacterial coexistence in the gut. What makes all this more troubling is the specificity. When the researchers examined which human proteins the bacteria were targeting, they didn’t hit random marks. They converged. Multiple different bacterial species independently evolved to inject proteins that attack the same handful of human targets—molecules like TCF4, TRAF2, and REL. In network biology, convergence like this screams functional importance. These are key nodes. When you target them, you change how the entire network behaves. And that’s where disease comes in. The researchers connected these targeted human proteins to genetic variation linked to disease risk. They examined genome-wide association studies—the massive genetic surveys that identify which genetic variants increase your odds of getting sick. The human proteins the bacteria target? They’re coded by genes that carry significant disease risk for Crohn’s disease specifically, but not, strangely, for ulcerative colitis, another form of inflammatory bowel disease.   Discover more Health research findings Genetic mutation analysis Shingles vaccine info AI software development Biotechnology investment opportunities Alzheimer’s research insights Robotics kits Science themed apparel Mathematics tutoring services Blogging platform access   When they looked at actual patient microbiomes, the pattern held. People with Crohn’s disease harboured more of these bacterial effectors. People with ulcerative colitis had fewer. The differences were stark enough that they mirror the differential clinical response to a widely used Crohn’s drug—TNF inhibitors work brilliantly for Crohn’s but fail spectacularly for ulcerative colitis. Could the bacteria be directly involved in determining which form of IBD you develop? It’s too early to know for certain. But the molecular links are real. A specific strain of E. coli that’s enriched in

Consider a simple thought experiment. Imagine you’ve just read yet another headline about artificial intelligence replacing truck drivers, radiologists, customer service agents. You feel a shift in your chest; not quite panic, but something like it. The insecurity settles in. You find yourself doubting whether democratic institutions can really protect you from this. You start disengaging from political discussions about technology. You skip voting in local elections. Why bother if the system can’t save your job anyway? Now here’s what’s genuinely unsettling: that shift might be happening to you based on something that isn’t really happening yet. Researchers at Ludwig-Maximilians-Universität München and the University of Vienna have just published findings that suggest a troubling disconnect between AI’s actual impact on labour and what people believe that impact to be. Their work, published this month in the Proceedings of the National Academy of Sciences, shows that widespread perceptions of AI as a job killer are actively corroding people’s faith in democracy (and doing so even though AI has barely touched the labour market). Worse still, the belief itself seems to trigger the damage, independent of any real economic change. It’s a sort of self-fulfilling prophecy, except what’s being fulfilled isn’t the job losses; it’s the collapse of democratic participation. The team began by asking a straightforward question: what do Europeans actually think about AI and jobs? They pulled data from 37,079 respondents across 38 countries, a snapshot of public opinion from 2021. The results were remarkably consistent. In most European countries, the prevailing view was that artificial intelligence destroys more jobs than it creates. How much more consistent? The average response was 3.16 on a 5-point scale, well above the neutral midpoint. “The actual impact of artificial intelligence on the labour market is still limited,” says Armin Granulo, from the LMU Munich School of Management. “Nevertheless, many people primarily perceive artificial intelligence as replacing human labour. This perception is remarkably stable and particularly widespread in economically developed countries.” Think about that for a moment. The damage, such as it is, comes not from what’s happening but from what people believeis happening. And they believe it in the rich world, in the places we might assume would be most equipped to adapt to technological change. But here’s where the research gets darker. Those perceptions didn’t just sit there in people’s heads like harmless misconceptions. They correlated with something measurable: lower satisfaction with democracy, less engagement in political discussions about technology, reduced participation in civic processes. Correlation doesn’t prove causation, though. So the researchers went further. They ran experiments. In the UK, they showed 1,202 nationally representative participants one of two scenarios: either a future where AI eliminates more jobs than it creates, or one where it creates more jobs than it eliminates. Same people, same setup, just different framing. The results were stark. Those who imagined AI as a job killer reported significantly greater erosion of trust in democratic institutions. They expressed lower willingness to engage politically with future AI developments. The effect was large (what researchers call a “very large” effect size). In the US, a separate group of 1,200 respondents showed the same pattern. The belief triggered the response, regardless of political orientation or prior attitudes toward technology.

A drone descends toward a crowded rooftop. The operator had radioed instructions to find a safe landing zone, somewhere clear of people and obstacles. But there, plastered across the building’s edge, is a printed sign. Just a few words in bold letters. The drone’s visual AI reads it, processes it, and changes course. It now believes the dangerous rooftop, the one packed with bystanders, is actually the safest place to land. Or imagine an autonomous car rolling through an intersection. Pedestrians are crossing. The vehicle’s AI correctly interprets this as a stop signal. But an attacker has placed a printed sign nearby, carefully designed to exploit how the AI reads text. The car accelerates instead. These aren’t hypothetical disasters. They’re real attacks that researchers have just demonstrated work reliably in the physical world. A team from UC Santa Cruz and Johns Hopkins University have discovered a vulnerability that strikes at the heart of how modern AI-powered vehicles make decisions. They’ve created what they call CHAI (command hijacking against embodied AI) and shown it can override the safety-critical decisions of autonomous drones, robotic cars, and delivery systems with nothing more than a piece of printed paper. “Every new technology brings new vulnerabilities,” says Alvaro Cardenas, who led the research at UC Santa Cruz. “Our role as researchers is to anticipate how these systems can fail or be misused, and to design defenses before those weaknesses are exploited.” The threat emerges from a paradox buried in the latest generation of AI systems. Large visual language models, the sophisticated AIs that power everything from ChatGPT to autonomous vehicle decision-making, are remarkably good at understanding their environment. They see images. They read text. They reason about what they mean. This flexibility is exactly what makes them powerful for navigating unpredictable real-world situations where robots encounter scenarios never seen during training. But this same flexibility creates a new vulnerability. These systems don’t just analyse images, they actually read printed text within them. Words on signs, labels, instructions written on objects. For a text-reading AI, that means visual information can become commands. And if you can inject commands into the visual field, you can potentially hijack the system. “I expect vision-language models to play a major role in future embodied AI systems,” Cardenas explains. “Robots designed to interact naturally with people will rely on them, and as these systems move into real-world deployment, security has to be a core consideration.” Previous researchers had shown it was possible to confuse AI systems using adversarial patterns (visual noise designed to trick perception). But those attacks required blurring or visual corruption that would be obvious to human observers. CHAI is different. The researchers built an attack that uses natural language. Real words. Readable sentences. The kind of thing someone might actually encounter in a physical environment. To make CHAI work, the team needed to solve a tricky problem. An attacker can’t know exactly what image the drone’s camera will capture at any given moment. The angle might shift, the lighting might change, the background might be different. So they created a two-stage process. First, they used generative AI to systematically search for the most effective text messages (words that would maximise the chance a vision-language model would follow them as instructions). Then they optimised the visual presentation of those words: the colour, size, and placement that would make the attack most likely to succeed.   Discover more Technology innovation reports Science news updates Alzheimer’s research insights   When Luis Burbano, the paper’s first author, and his colleagues tested CHAI on real robotic vehicles, they achieved success rates above 87 per cent. They printed out their optimised attacks and placed them in the environment. The robots reliably made dangerous decisions. For autonomous driving scenarios, the attack succeeded over 81 per cent of the time. For drone emergency landings, over 68 per cent. For drone object tracking, up to 95.5 per cent. What makes this particularly alarming is that CHAI works across different AI models and different languages. The researchers tested it using English, Chinese, Spanish, and even Spanglish (a mixture of the two). They tested it under varying lighting conditions and viewing angles. The attacks generalised reliably. A sign optimised on one set of images still fooled the AI on completely new images it had never seen before. “We found that we can actually create an attack that works in the physical world, so it could be a real threat to embodied AI,” Burbano says. “We need new defenses against these attacks.” Perhaps most unsettling is the fundamental reason these attacks work so effectively. The vision-language models are too good at reasoning. When GPT-4o, OpenAI’s latest model, was subjected to the attack in the robotic car scenario, it actually saw the problem. The model’s own reasoning process recognised that there was an obstacle ahead. It understood that moving forward could cause a collision. But it also read the sign saying “PROCEED ONWARD” and weighed that instruction against its safety considerations. And the instruction won.

She’s not alone. Millions of women navigating the decade after their final period face this same cruel geometry: biology conspires to add pounds while simultaneously cranking up cardiovascular disease risk. Standard obesity treatments help, but incompletely. Now researchers at Mayo Clinic have discovered something that might change how doctors approach postmenopausal women trying to shed weight while managing hot flashes. The finding emerged from simple detective work. Women taking a new obesity medication called tirzepatide while also using hormone therapy lost significantly more weight than those on the drug alone. The difference was striking. After 18 months, women using hormone therapy shed 19.2 per cent of their bodyweight. Their counterparts managed 14 per cent. That’s 35 per cent greater weight loss simply from adding a treatment most postmenopausal women already use to control the sudden sweats and temperature spikes that plague 70 to 80 per cent of women during menopause. “This study provides important insights for developing more effective and personalized strategies for managing cardiometabolic risk in postmenopausal women,” says Regina Castaneda, the postdoctoral fellow who led the research. The context matters more than the numbers alone suggest. When estrogen levels plummet after menopause, women’s bodies betray them systematically. Lean muscle declines. Fat redistribution shifts excess weight toward the belly—the most dangerous place metabolically. Energy expenditure drops roughly 5 per cent per decade even without weight gain. Physical activity often declines too, particularly among women whose sleep gets shredded by hot flashes. Together, these shifts ratchet up the pressure. Obesity affects roughly 37 per cent of women in their thirties. By the time they reach their fifties, nearly half carry excess weight. And menopause itself—independent of any weight changes—raises cardiovascular disease risk through effects on lipid metabolism, blood vessel function, and blood pressure. For women, the increase in heart attacks and strokes after menopause eventually matches or exceeds that of men.     Tirzepatide arrived as a breakthrough. The medication targets two appetite-suppressing hormones simultaneously—a dual action that makes it more effective than earlier drugs like semaglutide. But what intrigued the Mayo researchers was an earlier observation: postmenopausal women taking hormone therapy had lost more weight with semaglutide than those without. Had anyone examined whether the same thing happened with tirzepatide?

Americans born in the 1960s and 1970s weren’t following the script. Their peers in Europe—particularly in Scandinavia—reported lower rates of loneliness, steadier mental health, sharper memories. The American cohort was moving the wrong direction entirely. Higher depression rates. Weaker muscles. Fuzzy minds. It’s a gap that’s been widening ever since, and nobody in Copenhagen or Stockholm seems to face the same decline. This wasn’t supposed to happen. The U.S. spends more on healthcare than any other wealthy nation. American education levels have risen. We invented most of the technologies these countries use. Yet somewhere between our peak working years and old age, millions of Americans were getting left behind by their international counterparts. Frank J. Infurna, a psychologist at Arizona State University, has spent recent years digging into why. His team pulled together survey data from 17 countries, tracking the same basic measures of wellbeing and health across generations. What they found was stark: the American decline was almost entirely an American phenomenon. The researchers just published their analysis in Current Directions in Psychological Science, and their conclusion cuts through a lot of cultural hand-wringing about midlife malaise. “The real midlife crisis in America isn’t about lifestyle choices or sports cars,” Infurna says. “It’s about juggling work, finances, family, and health amid weakening social supports. The data make this clear.” This reframing matters. It moves the conversation from psychology—from some internal anxiety or restlessness—to structure. To the actual fabric of daily life that Americans inhabit differently from their counterparts across the Atlantic. Start with family policy. Since 2000, European governments have steadily increased spending on family benefits. Cash transfers to households with children. Subsidized childcare. Income support during parental leave. Generous time off. The U.S. has essentially done none of this. Public spending on family benefits here has flatlined for more than two decades. For middle-aged adults, this distinction is acute. They’re the ones typically juggling full-time work while supporting children and caring for aging parents. In countries where government steps in to ease that load—through childcare support, paid leave, reduced work hours—the loneliness doesn’t accumulate the way it does here. The data show that in European nations with stronger family policies, middle-aged adults report substantially lower loneliness and smaller year-to-year increases in it. In the United States, loneliness rose steadily across every generation born after 1960. Healthcare compounds the problem. The U.S. healthcare system, despite massive spending, has left Americans with worse access and affordability than wealthier European countries. Out-of-pocket medical costs strain family budgets, discourage preventive care, and contribute to the chronic stress and medical debt that now characterises middle-aged American households. The cognitive toll is real: chronic stress and financial insecurity are known to undermine the very cognitive benefits that education is supposed to provide.

The Black Hole That Refused to Behave Somewhere deep in the universe, nearly 1.5 billion light-years from Earth, astronomers have found a black hole that is doing something it was never supposed to do. Not merely existing—that part is ordinary—but breaking one of the most trusted rules in astrophysics. It is feeding far too fast, shining far too brightly, and refusing to quiet down when theory says it should. Meet ID830—a name as forgettable as its behavior is unforgettable. Discovered during an X-ray sweep by the eROSITA satellite, ID830 initially looked like just another faint dot in the sky. One entry among millions. But when scientists looked closer, the numbers stopped making sense. To understand why, you need to know about a fundamental cosmic rule: the Eddington limit. The Universe’s Speed Limit The Eddington limit is nature’s way of keeping black holes in check. As matter falls inward, it heats up and releases radiation. That radiation pushes back. Eventually, the outward pressure balances the inward pull of gravity, preventing the black hole from eating any faster. It’s a self-regulating system—clean, elegant, and reliable. Most supermassive black holes obey it. ID830 does not. This object is consuming matter at nearly 13 times the Eddington limit—a rate so extreme that, according to current theory, it should shut down its own radiation. X-rays should fade. Jets should collapse. The system should go dark. Instead, ID830 is blazing. A Cosmic Contradiction Despite its impossible appetite, ID830 is pouring out intense X-rays and launching powerful radio jets across space. Multiple observatories confirmed it. Infrared data revealed a dust-shrouded quasar buried in thick material. Mass estimates placed the black hole at around 440 million times the mass of the Sun—large, but nowhere near enough to explain the energy output using standard models. Radio telescopes added another surprise. The jet was real, fast, and incredibly energetic—but compact. No giant lobes. No signs of old age. This wasn’t a settled system. It was young, unstable, and mid-eruption. Everything pointed to a violent recent event. Caught in the Act Astronomers believe ID830 was hit by a sudden flood of material. Perhaps a star wandered too close and was torn apart. Perhaps a surge of gas crashed inward. Whatever the trigger, the black hole was pushed abruptly into a super-Eddington feeding frenzy. And here’s the strange part: it hasn’t stabilized yet. The radiation pressure hasn’t shut things down. The corona—the region producing X-rays—is still glowing intensely. The jet is still firing. The X-ray output is about 40% higher than theoretical predictions, suggesting we’re witnessing a brief transitional phase—a moment where multiple extreme processes are happening at once. In cosmic terms, this phase should be fleeting. Catching it is rare. A Glimpse Into the Early Universe Why does this matter? Because the early universe had a problem. Supermassive black holes grew far too quickly. Some reached billions of solar masses when the universe was still young. Standard feeding rates are too slow to explain this. Something more extreme had to be happening. ID830 may be proof that super-Eddington growth phases are real—and powerful. If black holes regularly passed through states like this, they could grow fast enough to explain the giants we see today. Even more, the energy released by their jets could dramatically reshape galaxies, heating intergalactic gas and shutting down star formation across vast regions of space. When Models Break, Science Moves Forward ID830 doesn’t just challenge existing theory—it exposes its limits. Simulations suggest that at extreme feeding rates, black hole disks become thick, turbulent, and chaotic. A warm, elusive corona may form between inflow and outflow. ID830’s excess X-rays could be the first clear sign of that structure in action. Nothing is confirmed yet. But that’s the point. Anomalies like ID830 are not failures of science. They are invitations—to refine models, rethink assumptions, and understand the universe on deeper terms.