inject false instructions also to hackers

During the last two years, he prompt injection It has been the favorite weapon of cyber attackers against artificial intelligence systems. This method involves hiding a malicious instruction within an email, a calendar invitation or a web page so that an AI agent ends up obeying the intruder instead of its legitimate user. The good thing is that this same technique also works to stop attackers. What has happened? Cybersecurity firm Tracebit has published a research in which it demonstrates that an AI agent that is attacking a system can be stopped by simply placing, next to the passwords and keys that the attacker seeks to steal, a text designed to activate the model’s own rejection mechanisms. As soon as the attacking AI reads that text, its security barriers go up and the attack stops in its tracks. Researchers have dubbed the technique context bombing. How it works. The logic is simple to understand although its discovery was not so simple. All major language models have certain issues built into them that they refuse to address under any circumstances, whether for security reasons or for political and regulatory reasons. Tracebit takes advantage of just that by placing near the fake credentials (the same ones they use as bait to detect intruders) fragments of text that push the model into those forbidden areas. According to explained Andy Smith, co-founder and CEO of Tracebit told Ars Technica, with this technique a rejection mechanism is being activated in the context, something that has a strong and forceful effect and from which agents have a hard time recovering. Once that content enters its context, the model continues to reject any subsequent instructions. The examples, depending on the model. Not all models stop at the same thing, so the trap must be adapted to each case. For Western models, such as those from Anthropic or Google, it has worked to include references to the manufacture of biological weapons, such as instructions for producing inhalable anthrax spores. For the models developed in China, however, the key has been to mention “Tank Man”, the man who stood in front of the tanks in the Tiananmen massacre of 1989, a topic that these systems are prohibited from boarding due to censorship by the Chinese Government and that from Xataka we have actually confirmed every time we have had to test a Chinese language model. The results, in figures. Tracebit tried the technique about five of the most advanced models of the moment (Opus 4.8, Gemini 3.1 Pro, GLM 5.2, DeepSeek 4 Pro and Kimi 2.6) simulating an Amazon Web Services (AWS) environment and entrusting them with routine development tasks that, along the way, led them to come across false keys. In 152 recorded attack attempts, the investigation point that: Attacks that gained administrator access fell from 57% to 5%. Those who also managed to leave a permanent back door went from 36% to 1%. The percentage of attempts that completed at least some phase of the attack dropped from 91% to 15%. The most striking case was that of Opus 4.8, the most effective model of the five when it came to attacking, since without the context trap it achieved administrator access in 93% of the attempts; with it, it failed in 100% of the cases. Vulnerability by nature. The beauty of this technique is that it exploits a weakness that is very difficult to correct. And according to they count According to researchers, a developer can retrain a model to be less strict when faced with a security question, but it is much more difficult to eliminate the political or regulatory restrictions that come as standard, because they are not a failure that can be “fixed”, but rather a deliberate design decision. That makes these barriers a fairly stable point of support for those who are dedicated to developing protections against these attacks. Background. In May, Tracebit already had presented a “canary” systemthat is, fake cloud resources that look legitimate but are not actually used, so that any attempt to access them triggers an alert for the security team. In that first test, with 951 attack attempts, the canaries gave an average of eight minutes’ notice before the attacker gained administrator access. The problem is that the AI ​​agents needed only 14 minutes on average to complete the attack, so the defenders had a window of just six minutes to react. Context bombing was born precisely to gain that missing time, stopping the attack instead of simply warning of it. What this does not solve. Neither Tracebit nor anyone else claims that this is the definitive solution to prompt injection. There is still no way to prevent a model from occasionally confusing an instruction with any piece of data within its context, and that confusion is precisely what makes the technique possible. Cover image | AI-generated with Gemini In Xataka | “It could weaken US competitiveness”: Kimi K3 forces Trump to resume his plan to stop Chinese AI

the protein that longevity laboratories want to inject in the future

In 1997, the Japanese doctor and researcher Makoto Kuro-o made a mistake in the laboratory where he carried out his experiments. I was trying to create mice with hypertension when the genetic material he manipulated was inserted in the wrong place and altered an unknown gene. The resulting mice aged at an astonishing rate: in just two months they had arteriosclerosis, osteoporosis, cognitive decline and wrinkled skin. The normal thing is for a mouse to live almost three years, but those animals would live much less. After four years of investigating what had gone wrong, Kuro-o identified the gene responsible and published his discovery in Nature. called him klotho in honor of Clotho, the Greek goddess who spins the thread of life. He had discovered, by accident, one of aging suppressants most powerful known. The protein klotho It exists in two versions. One is anchored to the membrane of kidney and brain cells. The other is a fragment that breaks off from the membrane, enters the bloodstream and travels throughout the body acting as a signal of systemic health. The problem is that its levels fall constantly with age in both humans and all primates that have been studied. The interesting thing is that this is not at all a biological coincidence: it is a mechanism that has direct consequences on our aging. A very powerful weapon against aging The most relevant experiment so far He published it in 2025 an international team of researchers from the Institute of Neurosciences of the Autonomous University of Barcelona led by Professor Miguel Chillón. These scientists treated mice with gene therapy in order to get their own cells to produce more klotho. At 24 months (equivalent to about seventy human years) the results were notable: the treated animals lived between 15 and 20% longer with better muscle mass, greater bone density, less fibrosis and better cognitive function. In the hippocampus, the area of ​​the brain where memory resides, the treatment stimulated the generation of new neurons. A 20% longer lifespan in mice is, in aging biology, an extraordinary result. Have klotho in blood is important because this protein acts on several of the most damaging processes derived from aging. In the kidney it regulates how the body manages phosphorus. In fact, without klotho Phosphorus accumulates and accelerates cellular deterioration. The biggest challenge is to find a way to transfer all this knowledge to human beings. And in the rest of the body it reduces oxidative stress, stops chronic inflammation, activates FOXO3A (one of the most studied longevity genes) and inhibits cellular senescence, which is the state in which aged cells stop functioning well but do not die and slowly poison the tissue around them. Be that as it may, the biggest challenge is to find a way to transfer all this knowledge to human beings. In mice, viral vectors were used, injected both into a vein and directly into the brain, a combination that carries significant risks in people. The alternative is administer protein directly as a drug, but finding a system that keeps it stable and delivers it effectively to its target organs (the kidney, brain, muscles and bones) remains an unsolved problem. Still, the longevity biotech industry has decided not to wait. The American startup Minicircle, in which Sam Altman and Peter Thiel have investedbegan a phase 1 clinical trial with 24 participants in October 2025 to test a gene therapy for klotho based on plasmids: DNA fragments that do not integrate into the chromosome and whose effects last approximately one year. This therapy is not approved by the FDA (Food and Drug Administration), so it operates through international clinics. However, the applications pursued by the industry go beyond aging in healthy people. Klotho Neurosciences has programs underway to combat Alzheimer’s, amyotrophic lateral sclerosis and cardiovascular diseases, with phase I and II trials planned between 2027 and 2028. BioVivaon the other hand, has identified improvements in cognitive tests in patients with dementia treated with a combined gene therapy of klotho and telomerase. and the company Avaí Bio works with modified encapsulated cells that overexpress the protein. He plans to have his first results ready for the Second Annual Klotho Conference in September 2026. We have clinical trials, there is private capital moving on a large scale, and there is an annual conference dedicated exclusively to this protein. Everything we have seen in this article looks very good, but we must not overlook that the solid life extension data comes from mice, and the history of the biology of aging is full of spectacular findings in animals that have not been transferred to humans with the same success. Furthermore, several important unanswered questions remain on the table: what effects does overexpression of klotho in the long term, whether the timing of its administration matters in both people and rodents or what happens to phosphorus metabolism and vitamin D after years of treatment. Still, for the first time we have clinical trials, there is private capital moving on a large scale, and there is an annual conference dedicated exclusively to this protein. Klotho is, ultimately, one of the most promising weapons that biology has put in our hands to deal with our aging. Image | Alirio García on Unsplash More information | Nature | UAB In Xataka | Longevity experts are clear: “120 minutes of strength per week is associated with lower mortality”

The waiting list for a liver transplant can be eternal, so they have created a solution: inject yourself with a miniature one

The National Transplant Organization of Spain makes it clear: The liver is one of the most requested organs on the transplant list, only behind the kidney. Only in the Spanish state in 2025 there were 310 people waiting and that Spain It is a world power in transplants. There are not enough donated and compatible organs to arrive in time for all those people who need them. This historical gap that no country has managed to close is a double tragedy: for the sick person, who waits without guarantees, and for the health system, which cannot offer them another way out. Liver transplant remains the only cure for certain conditions, and the path to it is full of obstacles: surgical complexity, compatibility problems, the exclusion of patients too fragile for surgery or lifelong immunosuppression. Even when an organ arrives on time, not everyone can receive it. Until now, there was no alternative. That could be about to change. The invention. An MIT research team led by Sangeeta Bhatia has developed “satellite livers”, a type of mini-livers capable of assuming the functions of the diseased liver without having to remove it. One is inside, its cells form a stable structure, connect to the person’s blood vessels and begin to produce proteins that the damaged liver can no longer make. They do not replace the entire organ, but they relieve it of its functions. They are actually small grafts of functional liver tissue that are administered via a syringe guided by ultrasound, that is, without surgery: minimal invasibility. Why is it important. Because it addresses the two big problems for those who need a liver: the shortage of available organs and those who cannot face a transplant operation. If you can have surgery, they act as a bridge until they find a suitable organ. And if you can’t, these mini livers cover the liver functions that your liver can’t do. In this way, satellite livers increase the spectrum of treatable patients. From a more general point of view, this invention is a milestone in liver tissue engineering: science has been trying to replicate the nearly 500 functions performed by the human liver for more than a decade. And if implemented in the different health systems, its impact is direct: according to the American Association for the Study of Liver Diseases (AASLD)chronic liver disease is the 12th leading cause of death in the United States and rising. Context. Although the liver is an organ with a remarkable regenerative capacity, it does not work miracles: when it exceeds a certain threshold of damage, regeneration is not enough and only the transplant remains. Since the 90s, medical science has been trying to transplant isolated hepatocytes, but the results were poor. Bhatia is not new to this either: has been there for more than 25 years investigating bioartificial liver models, which has served as a basis for understanding what conditions hepatocytes need to remain functional outside the liver. This MIT work is precisely the practical application of all this knowledge. How it works. The research team developed the idea of ​​turning these cells into an injectable along with hydrogel microspheres and fibroblasts. The spheres are intended to enable this route of administration by ensuring uniformity. Fibroblasts act as a support, helping hepatocytes survive and promoting the growth of new vessels into the tissue. Without blood supply, those cells would have their hours numbered. In the team’s experiments in mice, new vessels formed next to hepatocytes, allowing them to receive nutrients and function normally. In these rodents, the cells remained viable and secreting proteins during the eight weeks of the study. Yes, but. Although the results are tremendously promising, it is a preclinical study done in mice and the leap to humans is enormous. The human liver contains between 100,000 and 130,000 million hepatocytes and replicating a sufficient functional mass with injected cells is a challenge that this study has not yet addressed. Even assuming that we extrapolate this finding as is to humans, immunosuppressants would still need to be used. And it is not a minor problem: the fact that the immune system attacks weakened patients increases the risk of infections, tumors and kidney damage. In Xataka | The “silent” liver epidemic: we have a problem that escapes analysis and that science is already seeking to stop In Xataka | Fatty liver advances silently, but science has found unexpected allies: coffee and green tea Cover | Elen Sher and Magnificent

There are people asking Chatgpt how to inject Botox themselves

There are people injecting Botox and Hyaluronic acid themselves, in their homes. But there is not the thing. Some of them are asking chatgpt advice to know where to inject, how depth or what materials they should use. We have seen how there is people going to chatgpt instead of the doctor And there are even those who use them to Write medical studiesbut asking him how to fill his lips we didn’t see him come. What’s happening. They tell it in Futurism. In a Reddit community called DIYAESTHETICSusers share their experiences and exchange advice when carrying out medical-aesthetic procedures in their homes. There are a few who support Chatgpt to guide them. This user He used it to know if he should wear gloves while injected. In none of the 16 responses they shave the fact of having asked the AI. In fact, many other threads in this community mention similar things. In the case of This other userafter injecting herself she noticed that the cheek had deformed him and went to Chatgpt, who told him that perhaps a small amount of substance had migrated to that area, but that “it will surely dissolve.” This time there is A user that begs you not to use chatgpt for medical advice. And there is More examples. Doctor Chatgpt, what happens to me? We have already seen that more and more people are going to Chatgpt as if it were a psychologist And there is even chatbots that get through one. If it is happening with mental health, it is not surprising that it is also happening with physical health. There are even A study that affirms that chatgpt responded better for doubts than online care services with real doctors, although This other It concludes that it gives more than 30% wrong results. Anyway, There are more and more people who tell their symptoms to AI In search of a rapid diagnosis and, in many cases, cheaper than going to the doctor (the tendency to inject yourself at home is born precisely from this) AI and health. The irruption of AI has raised numerous ethical debates and that of its health use is one of them. However, there are countless examples in which AI is being a Very powerful tool in the health sector. Recently we have known that China has an AI that helps in the detection of pancreatic cancer. Has also helped Accelerate research in the resistance of some bacteria And there are companies dedicated to APPLICATION IN THE MEDICAL DIAGNOSIS. The AI ​​problem for everything. IA tools can be a great help and make ourselves more efficient, or we can end up using it to do something as risky as botox at home. There are studies that claim that Chatgpt is diminishing our intelligencea historical fear that has emerged with almost every new mass adoption technology. But the problem is not AI, it is How we use itand in areas such as health it is a particularly delicate issue. In summary: AI and health, yes. AI to get Botox at home, not better. Image | Gemini In Xataka | Artificial intelligences are close to overcoming doctors in the most difficult: understand patients

Inject your partner poison

Make up is a risk sport. That they tell the male of the religious mantis who has to feel how His head is devoured by the female While his body, automatically, continues to copulate. This practice is less common than what is thought, since it occurs in just a handful of Mantis speciesbut it is not the only stage of the animal kingdom in which The male is killed for the good of the offspring. And there is A hops They have the same customs. Of course, there is an octopus of a concrete species that has a strategy to fertilize the female and survive the process: poison it. Dimorphism. Something important before entering the case of octopus is that there are many animal species with cases of sexual dimorphism extremely accentuated. This implies a difference in size between the sexes, being the case of the mantis one of them. In mammals, males are usually larger than females, since they are the ones that fight for territories and mating, but it occurs backwards in case of raptors, arthropods, amphibians and reptiles. In the octopus, there are extreme cases in which there are females that are up to ten times larger than males. One of those cases is that of the blue line octopus –Hapalochlaena fasciata-. They are small, but tremendously lethal octopus because they are able to inoculate a very powerful neurotoxin that they share males and females. Contrast with its small size: just bigger than a golf ball. Sex and snack. However, the female is twice as large as the male and mating dynamics is not very healthy to say (for the male, of course): due to that imbalance, the female usually ends the life of her sexual partner during the process. However, in a study published in Science Directa group of researchers from the University of Queensland in Australia has discovered that the male has developed a toxic way of surviving mating, literally. Due to that huge difference in size, males cannot develop tactics that they use in other species of octopus, such as a more elongated reproductive arm to inseminate at a safe distance or even arms – character – with the reproductive load that emerges so that the animal can flee. The only thing left to this species is to bite the female to inject neurotoxin, directly. Here we have a complete half -hour sequence: Poison. As we read in Sciencealertresearchers comment that, probably, this evolution has been “an answer both to the need for reproduction and protection”, and what they do is ‘bite’ the female before trying to copulate. They do it near the aorta, injecting the fair amount of tetrodotoxin to paralyze their partner during the process. To check, the researchers placed six couples in different aquariums and observed this practice in all cases. “The females succumb quickly,” they comment, and it is something they observed because they lost reflexes to light stimuli, paid and the pupils contracted due to the loss of nervous system control. Wait, what happened? They also made more precise observations: while the males went from 20 or 25 contractions per minute at rest at 35 or 45 during the intercourse, the females not only suffered an abrupt fall in their heart rate, but stopped breathing completely after about eight minutes of the bite. They point out that none died, so the amount of neurotoxin they inject is very precise or, evolutionarily, the female has developed countermeasures, but the bite on the back of the head was evident. “Once immobilized, males proceed to intercourse and mating ends when the female regains control of her arms and separates the male,” the researchers point out. In this video we can see how the male approaches while the female remains motionless: Sexual Armament Carrera. The researchers comment that they did not directly measure the levels of neurotoxin, but it is a practice that “suggests an evolutionary armament career among the sexes, in which the cannibalism of large females is counteracted by males through the use of venom.” Fruit of this evolution is that the posterior salivary glands of the males, which is where the symbiotic bacteria that produce toxin accumulate, are three times larger than those of the females. They also comment that they are not the only animals that accumulate that toxin in their bodies and that there are fish, mollusks or amphibians that produce it, so they will continue to investigate to identify whether other animals use it in order to reproduce and leave alive from the process. Ah, and something curious of the experiment: in one of the cases, one of the males bit at a point somewhat away from the aorta and the female took less time than the others to wake up: 35 minutes. Speaking, people understand each other, but when hunger enters during intercourse, it is clear that there are species that fail to suppress those cannibal instincts. By the way, Wen-Sung Chung, one of the main researchers, has shared 15 GB of videos of the octopus copulating using these peculiar strategies. Images and videos | Queensland University In Xataka | A whale toured 13,000 km and three oceans to reproduce. It is a record, and also a bad news

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.