trust

“We no longer trust US hyperscalers.”

by

several weeks ago, The US ordered Anthropic to suspend access to Fable 5. Shortly after the model was available again (although with changes), but this event served something else: to make tangible a threat that Europe had been talking about in the abstract for years. We spoke with Andreas Prins, global director of Sovereign Solutions at SUSE and expert in digital sovereignty, about what this episode reveals about European technological dependence, and why the political response continues to lag behind the problem. The trigger for Fable 5. The suspension of Fable 5 and Mythos, Anthropic’s most powerful models, marked a before and after in terms of digital sovereignty. According to Prins: “The most interesting movement that emerged from this is the awareness on the part of companies. We often talked about sovereignty, digital resilience and independence, but never to the level of a government effectively retiring software (…) people suddenly realized that dependency is real.” Prins, who works for one of the largest open source infrastructure companies, says conversations with companies have changed a lot after this incident. The same managers who once “ran to deploy Gemini, Anthropic, or global vendors like OpenAI” now realize that “I probably don’t need the newest, coolest model; what I need is a model I can control, running on my own premises or in my data center, with open source software I can audit and inspect.” Notice to sailors. However, the temporary suspension of these models did not cause havoc in European companies and institutions for a simple reason: they were very new models and there was no time to integrate them into critical processes. “If this suspension had occurred within three or five months, with companies operating their chatbots, customer service and decision-making engines based on these models. The impact would have been much greater,” warns Prins. In this sense, this event functioned more as “an early warning signal than a real-time crisis.” A question of trust. Trump’s obsession with taking over Greenland At the beginning of the year, relations between the US and Europe deteriorated rapidly. Europe began to realize something uncomfortable: The US was not the reliable partner I thought and that technological dependence could be a weapon of pressure. For Prins it is clear: “It is a question of trust and, if I evaluate digital sovereignty in Europe, the feeling is quite uniform: we no longer trust US providers or hyperscalers, and therefore we want to build our own alternatives.” Digital sovereignty is also a question of resilience, that is, what can happen to your business if software stops working tomorrow or if it stops receiving security patches. “Sovereignty is fundamentally a business risk assessment rather than a purely IT issue. You can take technological risks, as long as you do it consciously,” says Prins. A tangible example. To illustrate how risk perception is changing, Prins tells us the case of a company in the Netherlands that had its main environment in a data center in Frankfurt, managed by a hyperscaler, and also a data center below sea level in the Netherlands. The question was whether to maintain backup infrastructure in the Netherlands, where dam failure could cause flooding and cause serious damage. After a risk analysis, they came to a conclusion: “The risk of flooding from levee failure was lower than the risk of an American hyperscaler pulling the plug on European customers.” Wow, they trust more a structure which in many parts is already approaching a century of life. The current situation. Europe is already taking its first steps towards technological independence. It has been announced European Technological Sovereignty Package and there is also the Cloud and AI Development Actbut we are still in a very incipient phase and it is difficult to imagine a joint response of 27 countries, each with their own interests. Prins admits that “it will be difficult.” SUSE has worked with Denmark, France, Germany and the Netherlands in drafting recommendations to the EU, and their reading is that the disagreement is not in the objective, but in the execution since each one works with different tools. “To build a strong Europe we should unify these initiatives more and leave aside national particularities, although the market is large enough for several suppliers.” When asked about which countries are doing better, Prins avoids pointing out a single leader and prefers to talk about specializations: “Norway and Finland are taking giant steps in sovereignty applied to the defense industry”, in Germany there are states transforming their infrastructure on a large scale; in Spain, cites projects such as Penpot and the SUSE’s recent alliance with OpenChip. For its part, the United Kingdom “remains closer to the US for obvious reasons.” The importance of open source. Andreas Prins works in an open source company, but defends that its use goes beyond interest, but is a necessity: “to achieve the strictest levels of compliance required by the Cloud Act, the only viable way is to use open source,” he says. The advantages are auditability, forking capability, and exit speed, referring to “the ease of migrating to another system.” He contrasts this with how hyperscalers operate: “they capture customers with low entry costs and complex markets, whose long-term costs and dependencies are often underestimated.” When asked why this type of open initiatives have not caught on in Europe, he points out that “we should not underestimate the power of lobby of the big tech companies” and admits that there is also a visibility problem: “There are excellent open source alternatives in Europe, but they tend to be invisible. “I myself was surprised to see the potential we already have here.” Inaction is the worst scenario. When asked what would happen if Europe does not act in the coming years, Prins draws two scenarios. In the worst case “that geopolitical tensions calm temporarily, we put aside sovereignty ambitions and continue as before.” In the best case scenario (and the one he most trusts): “that figures such as Open Source Liaison … Read more

Someone used Opus 4.8 to analyze an MRI. What followed confirms that we cannot trust AI

by

A developer named Antoine Finkelstein had had a sore right shoulder for weeks. After visiting an orthopedist, an MRI was performed on the affected area, and according to the medical report, he had a grade III partial tear in the subscapularis tendon. Finkelstein then did something else: he passed the resonance to Claude Opus 4.8 to see what the AI ​​told him about that image, and the result was striking because according to that AI model his shoulder was “intact.” AI gives you clues. The developer suspected that the clinic was perhaps trying to cash in on his condition, so he requested the raw DICOM data from the MRI. What they gave him was 266 MB which he crossed with currently available AI models. First, of course, he made a quick consultation with ChatGPT and in it he detected potentially important negligence: the clinic had applied shock wave therapy, which is not recommended for tendinopathies without calcification. He had also been injected with Traumeel, a homeopathic product registered in Germany “without therapeutic indication.” Let’s see what Claude Opus 4.8 says. To try to get to the bottom of the matter, the user decided to turn the Anthropic model into a doctor to ask for a second opinion. After setting up the model on the Claude Code platform, he allowed the system to install the code packages needed to process the raw medical images he had been sent. After an hour of processing these images, the AI ​​model issued a surprising diagnosis: the tendon that human doctors detected as 50% torn was completely intact. I don’t trust. The result was so contradictory to the human diagnosis that Finkelstein wanted to go a little further and set up a blind arbitration system. He instructed Claude to deploy several independent subagents, combining AI images isolated from each other to avoid confirmation bias. The verdict of all those subagents was unanimous: there was no partial or total break, and everything suggested that the human specialists had exaggerated the diagnosis. But quantity is not quality. This article gave rise to an interesting debate on Hacker News in which some important reflections were raised. It is important to remember, for example, that although AI eliminates the cost of consultations, having more information is not equivalent to solving the problem. As I said For one user, the situation reminded him of a problem he had with his car. He asked three different workshops for a solution, and each one told him something, and one even recommended a repair that he knew was useless. “The solution to uncertain information is not more information, which is certainly what AI can providebut better information, and right now AI can’t provide that.” AI is too nice. There’s another problem here: big language models are meant to be nice and “nice.” They are in a sense echo chambers that want to keep us happy, so they are not designed to contradict us in a harsh way, which makes it easier to confirmation bias. If a user enters his suspicions in the prompt when asking the chatbot, the AI ​​tends to agree with him: we often see how he begins by answering with “You are absolutely right…”. The problem with answers to medical topics is that they can be very different in independent sessions, but since the tone is always convincing and confident, they can lead to more confusion than the initial one. The expert thinks. A professional radiologist participated in that conversation and provided expert insight. According to your criteriacurrent AI models remain mediocre at interpreting medical images due to the lack of massive public training databases. These data are protected by medical privacy laws, and at the moment this problem has a difficult solution, but that user explained that the latest models are already close in accuracy to that of a first or second year resident doctor. The theoretical threat to the radiology profession from AI is something we we have literally been talking for years: for now It doesn’t seem like something like this is close to happening.. Who is responsible. There is another big problem with AI: there is no one responsible if something goes wrong after applying a recommendation. It is true that human doctors can make mistakes and may have biases or even commercial incentives (selling treatments). However, the legal difference is fundamental: the medical system has a series of licenses, regulations and responsibility management that penalizes negligence. AI forces you to manage yourself in the face of uncertainty. The problem is simple: trust AI or not. In issues as delicate as this, it is proven that AI is still far from being a real substitute for human experts. Today’s medicine may be “commoditized,” but AI, no matter how cheap or attractive it may seem, does not yet have the precision that would be needed for certain areas. As Finkelstein himself concluded, “I can’t know if I can trust the AI, so I’m in a kind of limbo in which I either try my luck with another doctor, or I wait and see if my shoulder improves with the rehabilitation I’m doing.” Image | Vitaly Gariev In Xataka | A team in Malaga has just developed a new medical AI. Your job: Help interpret MRIs, CT scans and medical images

CATL wants a battery as powerful as gasoline. And he will trust his plan: lithium-air

by

CATL has prepared a very interesting roadmap for us over the next few years. With an energy transition increasingly accentuated in the automotive industry, there are several battery technologies that will fight for permanence in the next decade. Wu Kai, chief scientist of CATL and academician of the Chinese Academy of Engineering, advertisement At the Equipment and Energy Forum 2026, the company has identified lithium-air technology as the strategic front where the next great global battery battle will be fought. It is the first time that CATL makes this bet officially public. Why this ad matters. CATL controls 47% of the global electric vehicle battery market, according to April 2026 datawhich means we are talking about the world’s largest battery manufacturer by market share. The company has also accumulated five consecutive years as a leader in global energy storage, with a share of 30.4% in 2025. So, when its chief scientist points out a technology as the battlefield of the future, the industry listens. What exactly is a lithium-air battery. Unlike conventional lithium ion batteries, which use heavy metal compounds (nickel, cobalt, manganese) to house lithium ions, lithium-air batteries dispense with that solid cathode and replace these materials with oxygen taken directly from ambient air. The anode is pure metallic lithium. The result is a lighter system with an open architecture, which has led researchers to call them “breathable batteries”. Without so much dead weight inside the cell, the potential energy density skyrockets. The numbers. The theoretical energy density of this technology reaches 12,000 Wh/kg, a figure comparable to that of gasoline, which is around 13,000 Wh/kg. The lithium ion batteries that equip electric cars today offer between 250 and 270 Wh/kg. Solid-state batteries, considered the next big leap, aim for about 500 Wh/kg. The lithium-air prototypes already developed in the laboratory have exceeded 1,200 Wh/kg, more than four times the capacity of current batteries. If this technology were commercialized, we would be talking about electric cars with ranges of more than 1,600 kilometers on a single charge. A problem that comes from the 70s. The lithium-air battery concept is not new. And just as share CarNewsChina, its theoretical foundations were laid out in the 1970s. The problem is that taking it from theory to practice has proven extraordinarily difficult. The cells are very sensitive to humidity and carbon dioxide present in the air, which causes rapid degradation. Added to this are problems with catalyst stability and a very short useful life. But there is real progress. In 2024, a joint team from the University of Illinois Chicago, Argonne National Laboratory and California State University Northridge managed to demonstrate a lithium-air battery capable of exceeding 700 charge cycles in an environment similar to real air. A year later, in 2025, Argonne National Laboratory and the Illinois Institute of Technology developed a prototype that reached 1,200 Wh/kg with a life of 1,000 cycles at room temperature. According to collect CarNewsChina, this design is not expected to be ready for use in vehicles before 2030. The key to the breakthrough was, among other things, replacing liquid electrolytes (which are flammable) with a solid matrix composed of a ceramic polymer with lithium-rich nanoparticles, which stabilizes the cell during high-energy cycles. How does this fit into CATL’s strategy. The company already has experience in converting alternative technologies into market products. An example is sodium-ion batteries, which were proposed by the company in 2020 and This same year they are already being mass producedinstalled in models such as the GAC Aion UT, the Changan Oshan 520 and vehicles from Geely, Chery and FAW. According to explained Kai in the forum, the company’s strategy is planned in the short term to offer mature technologies to meet current demand; in the medium term, solid state batteries to improve the experience in premium vehicles; and in the long term, lithium-air with the intention of exploring the physical limits of energy storage. Between the lines. Betting on lithium-air now is not waiting for a product for next year. Just like points out Gasgoo, for large companies, investing in these frontier technologies serves above all to accumulate patents, secure strategic positions and build technical reserves, not to generate short-term income. It is something like a move to avoid surprises in case another company decides to announce a disruptive technology. Cover image | CATL In Xataka | Peugeot, on PureTech engines: “We recognize that we have done things wrong”

The engineers who worked at FSD do not trust their own creation

by

Elon Musk has been ensuring for a decade that full autonomous driving is just around the corner. Although the company has advanced in its driving assistance systems, a Reuters investigation reveals something worrying. Several people who worked on this project have denounced that the technology continues to suffer from basic and dangerous errors, and they confess that they would not get into a Tesla autonomous car for the world. what has happened. In this investigation, Reuters had the testimonies of nine “data taggers”—the people who train that Tesl AI system—as well as a software engineer who worked on the project. According to them, vehicles with these systems collide with animals, ignore the presence of school buses or accelerate in construction zones. One of the team’s veterans summed up everything in one sentence: “we have all seen the FSD fail.” Beware of public demos. Tesla has already launched robotaxi pilot programs in cities like Austin (Texas). Musk claims that his software is a generalized system that can be adapted to any city without high-precision maps, but these interviewees indicate that the operational reality is different. The trick. Tesla staff spent months recording videos and mapping the area of ​​Austin where the tests were to take place, and they spent hundreds of hours labeling curbs or road markings just to avoid problems during the demonstrations. According to these former employees, this level of intervention is unaffordable on a global scale. Comparing pears with apples. To maintain that the FSD system is up to ten times safer than human driving, Tesla uses a methodology criticized by experts. For example, he compares his cars (4.1 years old on average, modern safety systems) with the average American car, which is almost 13 years old. Phil Koopman, a professor at Carnegie Mellon University, explained that “It’s like saying my jet plane is faster than a World War II bomber.” The data reveals that if only accidents with airbag deployment are compared, Tesla’s advantage would not be 10 to 1, but 3 to 1, and even that figure is questionable. The controversial “Mad Max” mode. Internal videos have shown Tesla cars driving at speeds much higher than those allowed after the introduction of certain aggressive driving modes like the so-called “Mad Max”. Some of the employees who participated in the investigation reported cars traveling at almost 100 km/h in zones limited to 40 km/h. This aggressive driving is often treated as a low priority problem by its engineers, despite the risk it poses to road safety in these urban environments. Investigations in progress. The National Highway Traffic Safety Administration (NHTSA) currently has four open investigations into FSD and Autopilot. These cases include situations in which Tesla vehicles ignored red traffic lights or they turned directly into oncoming traffic. Fatal accidents that occurred are also being investigated in low visibility conditions —fog, sun glare—, and where the Tesla sensors, which are focused entirely on the use of cameras, have turned out to be insufficient. Where are the robotaxis? Almost a year after its launch in Austin, Tesla’s fleet of robotaxis it’s still tinyand consists of about 50 vehicles. It is also limited to very specific areas, and in cities like Dallas or Houston, users have complained that the cars do not drop them off at their exact destination. Besides, many of these vehicles They still have human drivers in the passenger seat who are there to avoid problems. It’s a reasonable practice, but it destroys the promise of full unattended autonomy that these vehicles offer. In Xataka | Elon Musk has come up with two names for Tesla’s self-driving taxi. And legally you can’t put any on it

The AI ​​industry fell in love with OpenAI, but doesn’t trust its CEO one bit

by

At OpenAI they see a future in which the work week should have four days. Not only that: every citizen should receive a share of the economic growth generated by AI. These are some of the proposals that the company has published yesterday with the aim of preparing us for the “age of intelligence.” And just the day they published that proposal full of good and reassuring intentions, a blow arrived for the CEO of OpenAI, Sam Altman. An investigation published in The New Yorker once again called into question his way of acting, highly criticized by experts and engineers who worked with him. The conclusion of all of them: better not trust Sam Altman. The arrival of the age of intelligence. What they call the “age of intelligence” will undoubtedly have a negative impact in some areas, but OpenAI proposes with their document to make changes that mitigate these problems. Among the most striking measures is the creation of a “public wealth fund” that will distribute dividends from AI directly among citizens, regardless of their employment status. Let the machines work (and pay us for it). They also suggest taxes on automated labor to finance social security, and also pilot projects of four-day work weeks without salary reduction. The proposal is striking and seeks, of course, to reassure citizens in the face of threats such as job loss that can be caused by the mass adoption of AI. The problem is that this proposal comes at a delicate moment for an OpenAI in the midst of a reputational crisis. Smokescreen? This optimistic proposal contrasts with the report published in The New Yorker and in which the authors interviewed more than 100 people “with first-hand knowledge of how Altman behaves in business.” And among them, rivals like Ilya Sutskever or above all Dario Amodei who founded their own startups. Both harshly criticized Altman. Sutskever accumulated internal documents and messages showing deception and manipulation. Amodei stated that the obstacle to AI security is Altman himself, who leaves that area in the background compared to the company’s ambition for personal power and excessive growth. For his former partners, Altman is not a visionary, but an actor with a calculated pose. Says one thing, does another. The scandal of dismissal and later return of Altman was due precisely to that attitude in which the council accused him of having “not been consistently frank in his communications.” It’s the same thing we’ve read on other occasions: Altman has a dual personality. In him, the pathological desire to be liked and accepted is mixed with a total lack of concern for the long-term consequences of his misdeeds. He tells his interlocutors what they want to hear, and then does what he really wanted from the beginning. It is something that, for example, Karen Hao narrates over and over again. in his book ‘Empire of AI’in which, it must be said, it erred in calculating the water consumption of data centers mentioned in its studies. In the report they mention how the well-known programmer Aaron Swartz met him before die in 2013 and commented about him even then that “he is a sociopath.” Public image is everything. The publication of the OpenAI document occurs at a particularly critical time for the company, which is involved in a reputational and strategic crisis. Anthropic has managed to become the darling of the AI ​​industry —without being much less perfect— and OpenAI has realized that it was experimenting with too many AI applications that were not profitable and now wants to refocus on what makes it profitable. The good intentions shown in the document try to get public opinion on their side just when the company plans its IPO. Learning from the past. Altman’s critics reveal that he is an expert at designing control mechanisms that go up in smoke. Support AI regulations (at least those that favor you) and publicly promotes ethics committees and alignment and security of the AI ​​that in reality later knocks down internally, at least according to those who work with it. It happened when he promised to allocate 20% of the computing capacity to the super-alignment team, and then actually gave up only between 1 and 2% of that capacity. Jan Leike, who was named co-leader of that team along with Sutskever, resigned in May 2024 indicating that “safety culture and processes have been relegated to the background compared to flashy products,” he explained in a thread in X. He ended up signing for Anthropic. Interested reviews. Although Altman’s career at the head of OpenAI –with what happened to the Pentagon as a recent example—reinforces the comments of those who criticize him, it must be remembered that competition in this industry is currently fierce. Many of those who participate in the report are direct rivals and therefore their criticism, veiled or not, is partly self-serving because it harms their competitor. In Xataka | There is a new generation of AI models at the doors and Anthropic has to sell them: “The biggest and smartest”

OpenAI says its agreement with the Pentagon is completely secure. His way of convincing us: “Trust us”

by

Don’t worry about anything, really. Trust us. Who says it is OpenAI, a company led by Sam Altman that has earned the reputation of saying one thing on one hand and doing another on the other. There are whole books written on that premise, and it is inevitable not to remember it now that this gigantic startup has signed a disturbing agreement. soap opera. OpenAI reached an agreement with the Department of Defense to integrate its AI models into government agencies, replacing Anthropic. They did so by indicating that they would impose requirements on the use of these models and would have red lines similar to those defended in Anthropic: no mass espionage, no development of autonomous weapons. That decision has cost Anthropic the contract with the DoDbut also has been tagged as a “risk to the supply chain.” Trust us. There are two problems here. The first, that OpenAI has never shown the contract that makes it clear that there are red lines to the use of GPT by the military. And the second and most serious, that according to OpenAI we do not need it because we only need to trust them. Altman himself tried to dispel doubts explaining that they had added amendments to the agreement to ensure that those red lines were not crossed. The wall of opacity. Despite promises of transparency, OpenAI refuses to publish the contract. The firm’s head of national security, Katrina Muligan, he came to affirm in that it does not feel “obliged” to share the legal language of the agreement. This has raised suspicions about what has really been signed behind the scenes. Holes everywhere. Brad Carson, who served as secretary of the US Army under Obama, indicated at The Intercept how Sam Altman’s legal language in his posts on X is suspect. The CEO of OpenAI mention for example that “the AI ​​system will not be intentionally used for domestic surveillance of US citizens.” That “intentionally” is, according to experts like Carson, a kind of blank check to allow data on American citizens to be captured while spying on foreigners “by accident” but systematically. As Carson explains, They are trying to confuse you with complicated legal terms that ordinary people think mean something completely different. But lawyers know what it means. And lawyers know that this is no protection. The human factor. The integration of OpenAI’s AI into DoD systems now falls under the direct supervision of Secretary of Defense Pet Hegseth and President Trump. This represents an ethical dilemma: the security of the system depends on the political will of figures who have traditionally had no problem eliminating restrictions on mass surveillance systems. Quo vadis, OpenAI. The 180º turn it’s clear for OpenAI. While in its beginnings the startup was defined With the message of creating AI systems “for the benefit of humanity” and prohibiting the military use of its technology, this agreement demonstrates that such premises no longer seem to exist. another bad sign. This way of acting by OpenAI has caused it to be openly criticized on networks, but there have also been internal problems. This is demonstrated by the fact that its director of robotics, Caitlin Kalinowski, has decided to resign from office over concerns about the company’s military negotiations. And an obvious question. The dispute between the Department of Defense and the Pentagon centered precisely on the fact that they did not want Anthropic to establish red lines. OpenAI claims to have established basically the same ones, so how is it possible that the DoD allows OpenAI to establish them when it has not allowed Anthropic to do so? It doesn’t seem to make any sense. What a mess. We are living a real soap opera with three protagonists. The US Department of Defense (DoD) – now renamed the Department of War –, the company Anthropic and its rival, OpenAI. The DoD, which used Anthropic’s AI for military operations, He demanded to be able to use it without restrictionsbut Dario Amodei, CEO of the startup, he flatly refused. That was the moment Sam Altman took advantage of to become the new ally of the DoDsomething that has been seen by many as opportunistic and morally reprehensible. Image | Xataka with Freepik In Xataka | The war between Anthropic and the Pentagon points to something terrifying: a new “Oppenheimer Moment”

We can no longer trust any image on the internet

by

In 2012, Hurricane Sandy devastated the Caribbean Sea and reached the coast of New York. There he left floods, power outages and spectacular photos. Of all of them there was one especially amazing which went viral, but there was a problem: it was false. She wasn’t the only one that slipped into networks. That image was just one more example of what we have seen before and after: great phenomena and events end up generating floods of content, some of which are not real. There are many reasons why people take advantage of these moments to spread false images, but at least before achieving credible images and videos was expensive. Only advanced users of applications like Photoshop or Final Cut/Premiere could achieve convincing results, but AI, as we know, has changed all that. We have been warning about this problem for some time: distinguishing between what is real and what is generated by AI it’s getting harder. and these days we have had the last great example of this trend. Anatomy of a deepfake The Kamchatka Peninsula, in the far east of Russia, has experienced a historic snow storm. The worst in decades, according to records, with snow levels exceeding two meters in various areas, according to Xinhua. Petropavlovsk-Kamchatsky, the administrative, industrial, and scientific center of Kamchatka Krai, has especially suffered these consequences, and residents of the region have spread images on networks of the one that already has been baptized like the “snow apocalypse.” Those images spread in news media and social networks and that they were real—often more “mundane” and much less spectacular— contrast with others that theoretically also showed the state of various points in the region but that are actually generated with AI. That video, for example, was shared a few days ago by Linus Ekenstam, an influencer who often shares news and reflections on AI. He republished that video and claimed that it was real, but soon several users indicated that the video was actually created by AI. Ekenstam argued that the theoretical AI error that it pointed out in the user was not such, and that where he lives there are poles near the streetlights. He therefore tried to defend that for him the video was real, but others suggested that it was not. The definitive test: a user linked to the theoretical original videowhich apparently originated in a TikTok account dedicated precisely to disseminating AI-generated content that seems real. The crucial thing about that fake video is that it is spectacular, but not overly spectacular. It is, to a certain extent, believable, and when the image and the camera movement itself is so convincing, it is difficult to think that “maybe it is generated by AI.” With this snow storm experienced in Kamchatka, unusual images have been shared on networks, much more typical of a dystopian Hollywood movie than a real natural phenomenon. A priori the images may even seem coherent, but a more detailed – and above all, more critical – examination makes it easier for us to realize that perhaps these images are not as real as they seem. In fact, the most striking images shared on social networks and that accumulate thousands of retweets and likes on X, for example, contrast with those published in traditional media, which tend to be as we said much less flashy and much more mundane. Spanish media such as OndaZero or OKDiario have published some images and videos generated by AI on their digital media or on their social media accounts without realizing that these videos actually had their origin in the aforementioned TikTok account which has managed to spread like wildfire. Debates about the possibility that certain images could be real have been frequent for example on Redditwhere users shared for example an amazing catch which when analyzed in detail seemed generated by AI. The avalanche of “citizen journalism”, which can be well-intentioned and very important at times, contrasts here with the role of the media, which has an enormous responsibility in acting as trusted sources of information. Even they (and we) can fall into the trap, and here once again The best thing is to start distrusting what we see on our screens, because it may be false content. The videos that appeared in some media such as SkyNews or in The Vanguard they combine with others that (at least, a priori) seem real, but that at this point also require rigorous examination. Our brain betrays us and technology knows it There are several well-studied psychological phenomena and cognitive biases that explain why we believed in fake news in the past and now the same thing happens to us again with deepfakes. It doesn’t matter if we know (or at least rationally suspect) that these images and videos are false: technology and especially AI precisely exploit these biases. Among them the following stand out: Confirmation bias: we believe what fits with what we already believe. Our brain does not seek truth as much as internal coherence, so if a piece of news reinforces our ideology, we lower the level of potential criticism, but if it contradicts it, we analyze it with a magnifying glass or directly discard it. The problem here is that AI can generate tailor-made content adjusted to each narrative. Illusory truth effect: here it happens that “if I have seen it many times, it will be true.” Repetition increases the feeling of truthfulness, not actual truthfulness, and it is something that, for example, social networks, machines for repeating hoaxes, make the most of. Again, AI facilitates the mass production of the same lie with minimal variations. We believe what we see: This is what some call perceptual realism. We trust too much in the visual, and hence the famous saying “a picture is worth a thousand words.” Images are processed much faster than text, and critical thinking comes after the emotional reaction, as you well argued Daniel Kanheman in his famous ‘Think fast, think slow’. Cognitive load: related … Read more

“You can’t trust your eyes to know what’s real anymore.” Instagram CEO announces that the feed is dead

by

That the Internet as we knew it no longer exists is not a surprise: it has been filled with search results generated by artificial intelligence and from ‘slop‘. The consequences are already visible: clicks have been reduced by halfwhich is catastrophic for the media. But not only the text is suffering from this barrage of AI that blurs everything: already We do not know how to distinguish if an image is real or notwe have gone from document our life on social networks to the era of influencer content favored by the algorithm to videos and images that are not real, but can pass as such. There are no longer four fingers that are worth it. Instagrammers, the feed is dead. And this is also going to take its toll on social networks. Adam Mosseri, CEO of Instagram, closed 2025 with a publication in the form of a presentation of 20 images where he reflected in depth on what is coming: “the era of infinite synthetic content”, the antithesis of a more personal Instagram that has been dead for years. For Mosseri, AI has turned the carefully maintained grid with its algorithm into something of the past: “Unless you are under 25 years old and use Instagram, you probably think of the app as a feed of square photos. The aesthetics are careful: a lot of makeup, skin softening, high-contrast photography, beautiful landscapes,” Mosseri’s sentence falls like a stone on this millennial, who still uses Instagram as a kind of photo album. “That feed is dead. People largely stopped sharing personal moments on the feed years ago.” Tap to go to the post In search of something real. Mosseri explains that now its users keep their contacts up to date on their personal lives with “improvised photos of unflattering shoes and poses” shared via DM. And this also affects content creators: the omnipresence of images made by AI is going to bring a change: goodbye to those pro-looking photographs in favor of a more real and improvised aesthetic: “Flattering images are cheap to produce and boring to consume. People want content that feels real.” In fact, the CEO of Instagram points to manufacturers, applicable to cameras and mobile phones, who he says are making a mistake by democratizing the ability to “look like a professional photographer from 2015.” Because RAW images with defects are still a sign of reality until AI is able to copy them. But what is real? The time has come to unlearn to believe what our eyes see, something we have been doing all our lives. Javier Lacort explained that our entire epistemology (ranging from court testimony to photo albums) is based on the fact that seeing is a way of knowing. If you see a tiger, there is a tiger. If you see a photo of a tiger, someone has been close to one. This no longer applies: the era of uncover organized fake news has made way for anyone with Nano Banana Pro can get such an absurdly realistic image with a basic prompt in just a few seconds. Now creating a deepfake is trivial. Adam Mosseri think equal. “For most of my life I was able to safely assume that photographs or videos were largely faithful captures of moments that actually happened. That’s clearly no longer the case, and it’s going to take years to adjust. We’re going to go from defaulting to assuming that what we see is real to starting from skepticism. To paying attention to who’s sharing something and why. This will be uncomfortable: we’re genetically predisposed to believe our eyes.” If you can’t beat them… The paradigm shift has already occurred, so now Instagram and other platforms have to adapt to this new reality: “we have to build the best creative tools. Label AI-generated content and verify authentic content. Show credibility signals about who is posting. Continue to improve the ranking of originality.” It is the apocalypse of what is a photo that we have been predicting for years. Focusing on Instagram, Mosseri talks about how “we like to complain about ‘AI junk content,’ but there is a lot of amazing content created with AI.” He doesn’t give concrete examples or talk about Meta tools to make this possible, but Meta has already added AI tools on Instagram and Facebook. Without going any further, his AI Studio allows you to create personalized chatbots to deal with your followers. New times, new identification measures. It is increasingly difficult to identify content in AI, so it proposes fingerprints and cryptographic signatures in cameras to identify real content, forgetting about labels or watermarks. In any case, it advocates greater transparency about who publishes on the platform and improve creativity so that its human users can compete with content made in AI. In Xataka | The future of the Internet is to be flooded with AI. And there are those who have already seen a business niche: content made by humans In Xataka | There is a generation working for free as a documentarian of their own life: they are not influencers but they act as if they were.

They no longer trust their own debt

by

Deutsche Bank and Morgan Stanley are looking for ways to protect themselves from the debt they have extended to build AI data centers, according to Ed Zitron’s latest report in which he makes a notable criticism of the boom of AI and the stock market in which debt and complacent analysis are inflating an unsustainable bubble, according to their analysis. Both banks are contemplating “synthetic risk transfers.” It is a mechanism that allows the credit exposure of loans to be sold to other investors while keeping the loans on their books. Deutsche Bank even is considering betting short against actions related to AI. Why is it important. These movements clearly show a certain distrust in the economic viability of the infrastructure they are financing. Morgan Stanley, Deutsche Bank, Goldman Sachs, JP Morgan and MUFG have participated in the world’s largest data center financing transactions, including various loans to CoreWeave and the stargate projectsbut now they are looking to reduce their exposure to those same assets. The figures. At least $178.5 billion in data center financing was closed in the United States alone in 2025, almost triple the amount in 2024. CoreWeave, one of the largest operators, carries $25 billion in debt on estimated revenues of $5.35 billion, losing hundreds of millions each quarter. The context. AI data centers are powered by a circular financing model: They sign contracts with their clients before having the physical infrastructure. They use these contracts as collateral to obtain bank debt. They buy NVIDIA GPUs and build facilities that take between one and three years to be operational. Only then do they start generating monthly income. If construction is delayed or the client cannot pay, the loan is up in the air. Between the lines. The banks that have fueled the bubble are now covering their backs. Yes, but. Banks argue that these hedges are normal risk management practices. The problem is that they are hedging themselves against loans that they themselves structured and approved, many of them to clients whose ability to pay is, at the very least, uncertain. CoreWeave has offered OpenAI net 360 payment terms (one year from invoice to settle), depending on your loan agreement. If OpenAI, which needs to raise $100 billion to continue operating, decides not to pay, CoreWeave automatically defaults on its credit obligations. And CoreWeave is probably the best-funded operator in the IT industry. neoclouds. The money trail. NVIDIA announced in October that would guarantee $860 million in lease obligations from a partner in exchange for warrantswith 470 million deposited in a guarantee account. CoreWeave’s third-quarter balance sheet includes a “non-current restricted cash” item of $477.5 million. NVIDIA also signed a 6.3 billion contract with CoreWeave to buy the capacity that CoreWeave fails to sell until 2032. Go deeper. The banks that are hedging their bets are the same ones that have funded most of the global AI infrastructure. They are not selling the risk of any loan, but the risk of data centers that may never turn on, or that if they do, will serve customers who burn billions without generating profits. When the financiers of boom show signs of having stopped believing in boomit is worth paying attention. In Xataka | We have reached a point where not even the CEOs of Google or Microsoft deny that we have an AI bubble Featured image | İsmail Enes Ayhan

how it works and how to avoid this scam to steal money by earning your trust

by

Let’s explain What is the Like Scama new type of online deception that already Police and Civil Guard have warned. It is a scam that is being given in instantaneous messaging applications such as Telegram and social networks, and that will lead you to steal money after earning your trust. We are going to start the article explaining the mechanics and the procedure of this deception with which thieves earn your trust before Timing you. Then we will give you A series of tips To avoid falling into the trap. How is this scam First, scammers are going to contact you through different platforms, from Telegram to social networks. There, they will propose to you perform simple online tasks in exchange for moneysomething apparently easy that will not take you long. These tasks are things like Give and receive likes on social networksfollow profiles, etc. In exchange for this, they promise you small economic amounts for your time. With this they will feed both your trust and your greed. The tempting of all this is that At first they may pay you In exchange for what you are doing, they will give you small amounts of money. That is when your trust will be gained. When they have already convinced that they can help you earn money with little effort, they will rise to “higher groups”, where they will propose make investments in exchange for a lot of money. When you make these investments, cybercounts will simply disappear keeping your money. In addition, they will also stop paying for any task. And what is worse, the personal data you have given them to make the first payments are also possible to use them to Open bank accounts in your name and get more money at your expense. In addition, there are times when they can ask for money in exchange for continuing with these methods of winning, money that will also take. How to avoid falling this scam The first thing you should always do is distrust any method to earn fast moneysince everyone is usually deception online. It is sweet to be able to win some euros with simple tasks such as giving likes in accounts, but it is also a very common deception. Besides, suspect you who are asking for money or perform major actions. It doesn’t matter if you have been paid something, they are still people who do not know, so if you easily ascend in their ranges to be able to do greater actions, you have to suspect. Another important thing is Never give bank or personal dataand if they ask you right away you must be alert. Finally, no one with good intentions will ask you for money to continue working, even if it is an alleged “bargain” job. You must also suspect est. In Xataka Basics | Scam of the false winner on Facebook: how this scam works when you participate in competitions and how to avoid it

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.