Ilya Sutskever is not just another researcher in the technology sector. His career is directly linked to some of the advances that allowed deep learning to become the foundation of much of today's artificial intelligence and, years later, to the creation of Safe Superintelligence Inc. During his studies at the University of Toronto, he worked alongside Geoffrey Hinton, one of the great pioneers of neural networks and winner of the 2024 Nobel Prize in Physics. Together with Hinton and Alex Krizhevsky, he participated in the development of AlexNet, a neural network that achieved extraordinary results in the 2012 ImageNet competition and is considered one of the major turning points in deep learning.<\/p>\n\n\n\n After that period, Sutskever worked at Google Brain and in 2015 left Google to become one of the co-founders of OpenAI. There, he took on the position of chief scientist and participated in research that would eventually contribute to the development of the GPT family. His work helped consolidate a strategy based on progressively increasing the number of parameters, data, and computing capacity used to train models. The experience acquired during those years would later be fundamental to the creation of Safe Superintelligence Inc., a company that precisely proposes to investigate what may come after the large AI models.<\/p>\n\n\n\n However, Sutskever's vision evolved over time. The researcher began paying increasing attention to the consequences that the emergence of much more intelligent and autonomous systems could have. At OpenAI, he came to lead efforts related to so-called superalignment, whose objective was to investigate how to control artificial intelligence systems that could eventually surpass human capabilities. Finally, Sutskever left OpenAI in May 2024 and shortly afterward launched Safe Superintelligence Inc., also known as SSI, together with Daniel Levy and Daniel Gross.<\/p>\n\n\n\n The creation of Safe Superintelligence Inc. thus represents a new stage in Sutskever's career. In contrast to the development of artificial intelligence products aimed at the general public, Safe Superintelligence Inc. has set out a much more specific mission: to investigate how to build a superintelligence that is safe. This difference helps explain why Safe Superintelligence Inc. has generated so much interest despite keeping much of its research secret and not yet having a public product comparable to the leading AI systems.<\/p>\n\n\n\n Safe Superintelligence Inc. was born with a philosophy very different from that of the large companies that currently dominate the artificial intelligence market. Its name, Safe Superintelligence Inc., practically summarizes its entire mission: to build safe superintelligence. The company maintains that it has a single objective and a single product, rather than attempting to simultaneously develop a wide range of commercial applications. This decision is intended to reduce the influence that user needs, product launches, or financial results could have on research that may require many years.<\/p>\n\n\n\n The difference is important because most major AI laboratories must combine scientific research and business. OpenAI, Anthropic, Google, and Meta develop frontier models, but at the same time they need to turn them into products capable of attracting users and generating revenue. Safe Superintelligence Inc. has attempted to isolate itself from that commercial pressure in order to focus exclusively on researching a superintelligence that is safe from the outset. The company believes that safety should not become a solution applied after building an extremely powerful AI, but rather should be part of the development process itself.<\/p>\n\n\n\n For the time being, Safe Superintelligence Inc. has not publicly presented a model comparable to ChatGPT, Claude, or Gemini. Nor has it explained in detail what architecture it is using, how it is training its systems, or when it expects to achieve its goal. This lack of information makes it impossible to determine from the outside how far the project has actually progressed. What is confirmed is that Safe Superintelligence Inc. has managed to bring together top-level researchers and attract extraordinary amounts of capital.<\/p>\n\n\n\n Safe Superintelligence Inc.'s ability to attract funding is one of the most surprising elements of its story. The company initially raised around $1 billion and subsequently obtained another $2 billion in new funding rounds. These transactions raised its valuation to approximately $32 billion, an extraordinary figure for a company that had still not launched a known commercial product. Its investors include major venture capital firms such as Andreessen Horowitz, DST Global, Greenoaks, and Sequoia Capital.<\/p>\n\n\n\n The explanation for this valuation lies mainly in the combination of scientific talent and technological potential. Investors are not betting on a chatbot that already has millions of users or on a business with established revenue. They are betting on the possibility that Sutskever and his team will find a new path toward considerably more advanced artificial intelligence systems. If such a discovery were to occur, the economic consequences could far exceed those of any current AI application.<\/p>\n\n\n\n But it is important to distinguish between a multimillion-dollar investment and a scientific demonstration. The fact that sophisticated investors trust Safe Superintelligence Inc. does not mean that the company has built a superintelligence or solved the alignment problem. The publicly available information does not allow either of those claims to be made. Safe Superintelligence Inc. remains, above all, an extraordinarily ambitious research bet.<\/p>\n\n\n\n The situation changed significantly on July 27, 2026, when NVIDIA and Safe Superintelligence Inc. announced a long-term strategic partnership. NVIDIA confirmed that it had made an investment in the company and that Safe Superintelligence Inc. would have access to the Vera Rubin computing platform. Sutskever's company says that this new infrastructure will allow it to increase its computing capacity by an order of magnitude. Reuters put the amount of NVIDIA's investment at $5 billion, although the companies' official announcement did not publicly specify that figure.<\/p>\n\n\n\n The importance of the agreement goes beyond the money. The development of frontier artificial intelligence models requires enormous amounts of computing capacity, especially when researchers intend to experiment with increasingly large and complex systems. NVIDIA has become one of the leading providers of this infrastructure, and its processors are fundamental to a large part of the AI industry. For Safe Superintelligence Inc., having access to a platform such as Vera Rubin means having considerably greater resources with which to test its ideas.<\/p>\n\n\n\n The partnership also includes technical collaboration between the two companies to advance NVIDIA's current and future computing platforms. This is particularly relevant because Sutskever has stated that his team has research that it considers promising enough to scale. NVIDIA's investment can be interpreted as a sign of confidence in Safe Superintelligence Inc.'s work, although it does not by itself demonstrate that the company has found the solution it is seeking. The true result will only be known when Safe Superintelligence Inc. publishes advances that can be independently evaluated.<\/p>\n\n\n\n This is one of the major unknowns surrounding Safe Superintelligence Inc. Sutskever has expressed on different occasions his interest in better understanding the mechanisms that make capabilities such as reasoning, learning, understanding, and generalization possible. His approach starts from a fundamental question: we still do not completely understand what produces intelligence in artificial systems. Therefore, Safe Superintelligence Inc. appears to be interested in discovering new principles that could make it possible to move beyond the limitations of current systems.<\/p>\n\n\n\n There is not enough information to state that Safe Superintelligence Inc. has completely abandoned large language models. Nor do we know whether its architecture will be radically different from those currently used by other laboratories. What does seem clear is that Sutskever does not consider simply increasing the size of models indefinitely to necessarily be the definitive answer. His experience with scaling gives him precisely a privileged perspective from which to study what might work and where its limits could be found.<\/p>\n\n\n\n This possibility is what makes Safe Superintelligence Inc. a particularly interesting company. While much of the industry is competing to create larger, faster, and more capable models, Sutskever's team could be looking for a different way to produce advanced artificial intelligence. If Safe Superintelligence Inc. were to find an effective method, its consequences could affect the entire sector. But as long as there are no public results, any specific description of that technology should be considered a hypothesis and not a demonstrated fact.<\/p>\n\n\n\n Superintelligence poses a problem that goes far beyond obtaining more accurate answers. In general terms, superintelligence refers to a hypothetical system capable of vastly surpassing human beings in practically all relevant intellectual capabilities. A machine of this type could excel simultaneously in programming, mathematics, scientific research, engineering, strategy, and problem-solving. Its ability to learn and develop new strategies could be far superior to that of current systems.<\/p>\n\n\n\n The problem arises when such a powerful intelligence pursues an objective in a way different from what its creators expected. A highly capable system could find solutions that its designers had not anticipated or interpret certain instructions in unexpected ways. In a limited AI, an error of this type can be corrected relatively easily, but the scenario changes if the machine possesses far superior capabilities and a high degree of autonomy. That is why alignment has become one of the central issues in research on advanced AI.<\/p>\n\n\n\n The idea behind Safe Superintelligence Inc. is to try to solve this problem before a superintelligence emerges. Instead of first building an extraordinarily powerful system and subsequently looking for mechanisms to control it, the company wants safety to be part of the research process itself. This also explains why Sutskever considers that alignment cannot be a simple set of filters placed on top of a model. The challenge is to ensure that the fundamental behavior of a much more advanced intelligence remains compatible with human objectives.<\/p>\n\n\n\n It is too early to know, but there are reasons to think that Safe Superintelligence Inc.'s project seeks to explore new possibilities. Large language models have demonstrated extraordinary capabilities, but they still present limitations in areas such as reliability, long-term planning, deep understanding, and consistency of reasoning. The fact that Sutskever has focused Safe Superintelligence Inc. on researching the fundamental mechanisms of intelligence suggests that he does not intend to simply produce a slightly better version of existing technologies. The company appears to be seeking a deeper leap in the way artificial intelligence systems are developed and understood.<\/p>\n\n\n\n This does not necessarily mean that Safe Superintelligence Inc. will abandon the techniques that have driven the revolution of recent years. Advances in AI usually occur through a combination of different methods, and computing power continues to be an essential resource. NVIDIA's investment and access to Vera Rubin demonstrate that computing capacity will continue to play an important role in Safe Superintelligence Inc.'s strategy. The question is what new principles could be combined with that capacity to achieve a qualitative leap.<\/p>\n\n\n\n If Safe Superintelligence Inc. manages to discover an architecture or training method that is significantly superior, the impact could be enormous. An advance of this kind would affect not only chatbots, but also scientific research, robotics, programming, medicine, and automation. An artificial intelligence capable of improving its own capabilities much more efficiently could accelerate innovation in numerous fields. Precisely for this reason, safety becomes even more important.<\/p>\n\n\n\n Working in secret can offer obvious advantages. A laboratory researching potentially revolutionary technology can protect its discoveries, prevent competitors from copying its methods, and allow its researchers to work without the constant pressure to publish commercial results. In the case of Safe Superintelligence Inc., this model also fits with its intention to remain outside the usual demands of a technology product company. However, secrecy has an inevitable downside: from the outside, it is very difficult to evaluate actual progress.<\/p>\n\n\n\n For now, much of the confidence placed in Safe Superintelligence Inc. comes from the reputation of its researchers and the financial backing of major investors. NVIDIA's involvement constitutes another important signal, particularly because of the size of the investment and the access granted to its infrastructure. But none of these circumstances amounts to scientific proof that Safe Superintelligence Inc. has found the solution to the problem of safe superintelligence. That demonstration will have to come through concrete and verifiable results.<\/p>\n\n\n\n The future of Safe Superintelligence Inc. will therefore depend on what happens in its laboratory over the coming years. The company may achieve a breakthrough that changes the course of artificial intelligence, but it is also possible that its hypotheses prove insufficient or that the problem is much more complex than it currently appears. The history of AI shows that some predictions have come true sooner than expected, while others have taken decades to materialize. That is why it is advisable to separate Safe Superintelligence Inc.'s potential from any claims about results that we do not yet know.<\/p>\n\n\n\n The importance of Safe Superintelligence Inc. currently does not lie in a product that we can download or try, but in the question that it has decided to make the center of its research. While much of the industry is focused on making AI models faster, smarter, and more useful, Safe Superintelligence Inc. is trying to anticipate a stage in which machines could reach capabilities far beyond those of today. Its founder, Ilya Sutskever, knows firsthand the potential of scaling artificial intelligence and is now dedicating his efforts to solving the problem that could become its greatest challenge: keeping an extremely powerful intelligence aligned.<\/p>\n\n\n\n The partnership with NVIDIA represents an important boost for that mission. Access to much greater computing capacity will allow Safe Superintelligence Inc. to experiment with and scale its research, although there is still no public evidence that the company has achieved superintelligence or discovered a definitive solution to alignment. Its enormous valuation and the backing of major investors do not guarantee scientific success either. What exists is an extraordinary bet based on the experience of a team that considers it possible to advance toward a new generation of artificial intelligence.<\/p>\n\n\n\n The true value of Safe Superintelligence Inc. will become known when its results can be examined and compared with the advances of other laboratories. If it manages to discover new methods for developing advanced artificial intelligence and keeping it safe, its impact could far exceed that of any current chatbot. If it does not succeed, the research could also provide important knowledge about the limits of machine learning and the problems that will need to be solved before building even more capable systems.<\/p>\n\n\n\n For now, Safe Superintelligence Inc. represents one of the most ambitious and mysterious bets in the race for artificial intelligence. Its strategy is to work away from the spotlight, concentrate resources on a single objective, and avoid presenting premature results. In a sector accustomed to measuring progress through new products and models, Sutskever's company is trying to play a different game: to discover what makes superior artificial intelligence possible and, before taking it too far, find a reliable way to keep it safe. If that bet succeeds, the next great AI revolution may not begin with a new chatbot, but with a scientific discovery capable of changing our relationship with machines forever.<\/p>\n\n\n\n If your company needs specialized support to address today's technological challenges, ITD Consulting offers consulting services and IT solutions tailored to the needs of each organization. To learn about its services and explore how they can help your business, you can write to<\/strong> info@itdconsulting.com<\/strong><\/a>. <\/strong><\/p>","protected":false},"excerpt":{"rendered":" Safe Superintelligence Inc.: La superinteligencia segura que Ilya Sutskever desarrolla lejos de OpenAI y Anthropic ITD Consulting<\/p>","protected":false},"author":5,"featured_media":36853,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[16],"tags":[459,89,561,68],"class_list":["post-36849","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-noticias","tag-459","tag-ia","tag-safe-superintelligence-inc","tag-seguridad"],"_links":{"self":[{"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/posts\/36849","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/users\/5"}],"replies":[{"embeddable":true,"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/comments?post=36849"}],"version-history":[{"count":1,"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/posts\/36849\/revisions"}],"predecessor-version":[{"id":36854,"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/posts\/36849\/revisions\/36854"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/media\/36853"}],"wp:attachment":[{"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/media?parent=36849"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/categories?post=36849"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/itdconsulting.com\/en\/wp-json\/wp\/v2\/tags?post=36849"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}![\"ITD](\"https:\/\/itdconsulting.com\/wp-content\/uploads\/2026\/08\/itd-consulting-backup-acronis-vds-vps-ciberseguridad-microsoft-365-safe-superintelligence-inc-1-1024x576.webp\")
<\/figure>\n\n\n\nWhat Is Safe Superintelligence Inc.?<\/strong><\/h2>\n\n\n\n
A Company Valued at Billions Without a Public Product<\/strong><\/h2>\n\n\n\n
NVIDIA Bets $5 Billion on Safe Superintelligence Inc.<\/strong><\/h2>\n\n\n\n
![\"ITD](\"https:\/\/itdconsulting.com\/wp-content\/uploads\/2026\/08\/itd-consulting-backup-acronis-vds-vps-ciberseguridad-microsoft-365-safe-superintelligence-inc-2-1024x576.webp\")
<\/figure>\n\n\n\nWhat Is Safe Superintelligence Inc. Really Researching?<\/strong><\/h2>\n\n\n\n
The Great Challenge of Safe Superintelligence<\/strong><\/h2>\n\n\n\n
Is Safe Superintelligence Inc. Preparing an Alternative to Large Language Models?<\/strong><\/h2>\n\n\n\n
The Secrecy of Safe Superintelligence Inc. Also Raises Questions<\/strong><\/h2>\n\n\n\n
![\"Safe](\"https:\/\/itdconsulting.com\/wp-content\/uploads\/2026\/08\/itd-consulting-backup-acronis-vds-vps-ciberseguridad-microsoft-365-safe-superintelligence-inc-3-1024x576.webp\")
<\/figure>\n\n\n\n