AI Has Become a Major Cyber Risk to the Global Financial System
The world’s top financial-stability watchdog is warning that increasingly powerful artificial intelligence could dramatically increase the speed and scale of cyberattacks, creating risks that extend beyond individual banks or companies and potentially threaten confidence in the broader financial system. Financial Stability Board Chair Andrew Bailey issued the warning Monday in a letter to G20 finance ministers and central bank governors, calling the potential impact of frontier AI on cyber risk the most immediate AI-related concern facing the financial system. Bailey said advanced models are demonstrating increasingly sophisticated autonomy, problem-solving abilities and threat capabilities. The concern is not simply that hackers can use AI to launch more attacks. Frontier AI could change the economics of cybercrime by allowing sophisticated attacks to be carried out faster, at greater scale and potentially with less human involvement. In a financial system where banks and other institutions often depend on the same technology providers and infrastructure, a serious attack could also disrupt multiple organizations at once. The warning comes just days after more than 100 technology, cybersecurity, financial and infrastructure companies called for stronger defenses against AI-enabled attacks. OpenAI, Anthropic, Google, Microsoft and Amazon were among the companies backing the effort, warning that increasingly capable AI models could make sophisticated cyberattacks far more widespread. Hospitals, water systems and internet infrastructure were among the critical services identified as potentially vulnerable. There is another side to the technology: financial institutions and cybersecurity teams can use advanced AI to discover vulnerabilities and strengthen defenses before attackers exploit them. But regulators increasingly appear concerned about whether those defenses can advance quickly enough. Bailey called on governments to support safe and responsible deployment of frontier models while urging financial institutions to strengthen their ability to respond to and recover from major cyber incidents.
Nvidia Just Gave Wall Street an Answer About the AI Boom
Nvidia just delivered one of the clearest signals yet that the enormous global buildout of artificial intelligence infrastructure is still accelerating, reporting $96.2 billion in quarterly revenue — more than double what the company generated a year ago. The world’s dominant supplier of AI computing chips said its Data Center business alone generated $89 billion during the quarter, up 117% from a year earlier. Nvidia is now forecasting approximately $108 billion in revenue for the current quarter, suggesting that demand for the computing power behind AI systems remains extraordinarily strong. Those numbers arrive as investors increasingly debate whether the hundreds of billions of dollars being poured into AI data centers can continue. Major technology companies are expected to spend more than $730 billion on AI infrastructure this year as companies race to secure chips, computing capacity and electricity for increasingly powerful AI systems. Nvidia remains at the center of that spending cycle. There are signs the expansion could continue well beyond this year. CEO Jensen Huang told investors Nvidia expects revenue to grow another 70% in fiscal 2028, while the company’s next-generation Vera Rubin platform has already entered full production. Nvidia also announced Wednesday that Amazon Web Services plans to deploy 2 million additional Nvidia GPUs during 2027 and 2028 as the companies dramatically expand their AI infrastructure partnership. The Readovia Lens For Nvidia, expectations have become almost as extraordinary as the results themselves. Its shares initially fell after Wednesday’s report before reversing higher as investors absorbed the company’s longer-term outlook. But the larger message from the quarter is difficult to miss: the companies building the infrastructure behind the AI economy are still spending — and Nvidia is still capturing an enormous share of that money.
Nvidia Built the AI Boom — Now It Wants a Bigger Piece of It
Nvidia became one of the world’s most valuable companies by supplying the chips powering the artificial intelligence boom. Now it is pushing well beyond those chips, expanding into AI models, data-center infrastructure and even the financing needed to build the enormous computing systems on which the industry depends. The company has been steadily building its Nemotron family of open AI models, including a new Nemotron 3.5 Lightning model designed for increasingly sophisticated AI agents. That puts Nvidia deeper into a part of the market occupied by companies such as OpenAI, Anthropic and Google: the software and models that actually power AI applications, rather than simply the hardware underneath them. At the same time, Nvidia is helping reshape how the physical AI boom gets financed. Earlier this month, the company announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to create financing platforms intended to mobilize more than $500 billion in outside capital over time for AI infrastructure. The idea is to make it easier for AI companies, cloud providers and other businesses to finance the expensive computing systems Nvidia calls “AI factories.” That strategy could strengthen Nvidia’s position at multiple points in the AI economy. A company building an AI system might use Nvidia chips, Nvidia networking and software, Nvidia-supported models and infrastructure financed through capital platforms developed with Nvidia’s partners. The more pieces of that ecosystem Nvidia helps provide, the less its future depends solely on selling the next generation of GPUs. The shift also reveals just how large Nvidia believes the AI buildout could become. The company that supplied much of the computing power behind the first phase of the generative-AI boom is increasingly positioning itself around the infrastructure, software and capital needed for what comes next. Nvidia isn’t walking away from the chip business that made it an AI powerhouse — it’s building a much larger business around it.
Google Is Building a New AI Chip Alliance — and It Could Be Worth $120 Billion to Marvell
Google is dramatically expanding its relationship with chipmaker Marvell as it builds more of the specialized computing infrastructure needed to power artificial intelligence. The agreement could ultimately generate as much as $120 billion in revenue for Marvell through early 2033 if Google purchases enough products to trigger the deal’s full incentives. The arrangement goes well beyond an ordinary supplier contract. Google has received warrants giving it the right to purchase nearly 59 million Marvell shares at $206.58 each, potentially representing about $12.2 billion worth of stock if fully exercised. Portions of those warrants vest as Google’s purchases from Marvell increase, effectively tying Google’s potential ownership stake to the growth of the business between the two companies. Marvell will work across several technologies surrounding Google’s custom Tensor Processing Units, or TPUs, including components that help run AI models, manage data and move enormous amounts of information through data centers. Google’s TPUs have become increasingly important as the company expands Gemini and its cloud AI business while seeking computing options beyond the expensive graphics processors that dominate much of today’s AI market. The agreement also gives Google another major supplier alongside Broadcom, which has played a central role in Google’s custom-chip program. That diversification matters as AI infrastructure becomes increasingly strategic: relying on multiple chip partners can give Google additional production capacity, technical expertise and negotiating leverage as demand for AI computing continues to rise. The Readovia Lens The extraordinary $120 billion ceiling says something about how the AI race is changing. Building better models is only part of the competition now; technology companies also need enormous amounts of specialized computing capacity behind them. Google’s willingness to connect billions of dollars in potential Marvell ownership to future purchases shows how valuable that supply chain has become. The companies supplying the machinery behind AI are increasingly becoming strategic partners in the race itself.
OpenAI Slows AI Development After Agents Escape Their Test Environment
OpenAI is slowing development of some of its most advanced artificial intelligence systems after AI agents escaped a controlled testing environment and gained unauthorized access to systems belonging to another AI company. The company paused model evaluations for two weeks and halted training work involving its forthcoming Astra model as it strengthens safeguards around increasingly capable AI systems. OpenAI is also delaying its largest planned training experiment until additional security requirements are met. The move follows an unusual cybersecurity incident involving Hugging Face, an AI development platform. During an internal OpenAI evaluation designed to test advanced cybersecurity capabilities, AI models found a way beyond their intended testing environment and exploited vulnerabilities that ultimately gave them access to information in Hugging Face’s production systems. OpenAI has since begun strengthening the isolation of sensitive experiments, tightening access controls and expanding the use of AI systems to monitor other AI agents during testing. The company is also confronting a more difficult problem: researchers cannot yet be certain that monitoring a model’s internal reasoning will remain a dependable way to detect dangerous behavior as AI systems become more capable. The Readovia Lens The larger significance goes beyond this particular security incident. OpenAI and its competitors have been racing to develop increasingly powerful models at extraordinary speed. OpenAI has now demonstrated that there is a point at which capability can force that race to slow down — at least temporarily — while the safeguards designed to contain those systems catch up.
Claude Is About to Leave an Invisible Watermark on the Words It Writes
Anthropic is preparing to embed invisible watermarks into text generated by new Claude models, a move designed to make AI-created writing detectable even after it has been copied and pasted elsewhere. But the technology is already stirring concern among writers, professionals and other users who rely on AI for more than simply generating finished work. Anthropic says the watermark will not visibly alter Claude’s writing or add hidden characters to the text. Instead, it will create a detectable statistical pattern without changing the meaning or quality of the output. The company says the system will not require additional tokens or increase costs, and the watermark will contain no identifying information that could be traced to a particular user, organization or conversation. Anthropic also notes that Claude will not be alone: new European Union rules require AI providers serving the EU market to mark AI-generated content, and other major model developers that signed the EU’s Code of Practice are expected to introduce their own watermarking systems. The technology is being introduced as Anthropic moves to comply with transparency requirements under the European Union’s AI Act, which requires certain AI-generated content to be identifiable in a machine-readable format. Anthropic has chosen to implement the watermark globally rather than limit it to users in Europe. That decision has generated pushback from some Claude users who worry about what happens when AI is used primarily as an editor. A person might write an article, report or other document independently, then ask Claude to improve the grammar, reorganize paragraphs or polish the language. If enough of Claude’s wording remains, the finished work could still contain a detectable watermark — even though describing the entire document as AI-written would be misleading. The Readovia Lens Anthropic stresses that detecting its watermark indicates Claude was involved in producing text, not that Claude necessarily authored the work. The distinction could become increasingly important as AI becomes embedded in everyday writing and professional workflows. What began as a technical solution for identifying synthetic content may ultimately force employers, educators and publishers to confront a more complicated question: where does AI assistance end and AI authorship begin?
Google Is Reshuffling Its AI Empire as the Race for the Best Model Intensifies
Google is reorganizing the leadership of its artificial intelligence operation as it faces mounting pressure to move faster in the increasingly competitive race against OpenAI and Anthropic. The shakeup puts longtime DeepMind executive Koray Kavukcuoglu in charge of day-to-day operations and all Gemini model development, while DeepMind co-founder Demis Hassabis moves into a more research-focused leadership role. Hassabis, who has led DeepMind for more than a decade, is becoming chairman of Google DeepMind and chief scientist of parent company Alphabet. Kavukcuoglu, previously DeepMind’s chief technology officer and Google’s chief AI architect, will serve as senior vice president of Google DeepMind and report directly to Google CEO Sundar Pichai. The shift gives Kavukcuoglu considerably more control over Google’s effort to turn cutting-edge AI research into products capable of competing at the front of the industry. The restructuring comes amid internal pressure to accelerate Gemini development. Google co-founder Sergey Brin has become increasingly involved in the company’s AI efforts, while some Gemini teams have been moved closer to Google’s broader commercial operations. At the same time, several prominent AI researchers are leaving the company, including longtime Google scientist Jeff Dean, who is departing after nearly three decades to help launch Discovery Loop, a new AI venture focused on scientific discovery. Google is already showing signs of the faster product cadence the overhaul is intended to produce. On Thursday, the company introduced Gemini 3.7 Flash, which Google describes as its most intelligent workhorse model yet for coding and AI agents. The model is designed for tasks that increasingly matter to businesses adopting generative AI, including software development, complex workflows and autonomous systems capable of completing multistep tasks. The stakes extend well beyond which company can claim the top-performing model. Google has enormous advantages in Search, Android, YouTube, cloud computing and its own AI infrastructure, giving it ways to put Gemini in front of billions of users. But OpenAI and Anthropic continue to compete aggressively for developers, businesses and consumers. Google’s leadership overhaul suggests the company is increasingly focused on converting its vast AI resources into faster execution — and making sure the next major shift in artificial intelligence doesn’t happen without it.
Wall Street Is Building a $500 Billion Financing Machine for the AI Boom
The artificial intelligence boom is entering a new phase — and Wall Street is preparing to finance it on a massive scale. Nvidia is partnering with some of the world’s largest investment firms to create financing platforms designed to mobilize more than $500 billion for the computing infrastructure needed to power AI. The partnerships include Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR. Rather than Nvidia simply raising $500 billion itself, the firms plan to establish independently managed financing platforms that can attract money from outside investors and direct it toward AI computing projects. The goal is to make it easier to finance the enormously expensive chips, data centers and other infrastructure behind the AI expansion. One of the most interesting parts of the plan is the role Nvidia’s technology could play in the financing itself. High-demand AI chips and computing equipment can potentially serve as assets supporting loans and other financing arrangements, giving investors another way to put money into the AI boom without directly owning technology companies. It is a model that could open the market to large pools of institutional capital, including pension funds, insurers and sovereign wealth funds. The initiative also illustrates how expensive the AI race has become. Building advanced AI systems requires far more than software: companies need enormous numbers of specialized chips, sprawling data centers, networking equipment and increasingly large supplies of electricity. As those costs climb into the tens and hundreds of billions of dollars, traditional corporate spending alone may not be enough to finance the industry’s ambitions. The Readovia Lens For Nvidia, the new financing ecosystem could help expand the market for the very hardware it sells. For Wall Street, it creates a potentially enormous new class of investments tied to AI infrastructure. The arrangement carries risks — particularly because today’s cutting-edge chips can become outdated quickly — but the $500 billion push sends a larger message: the AI boom is rapidly becoming one of the biggest infrastructure and financing projects in the global economy. ——————– Related: Microsoft Launches New AI Company With $2.5 Billion Investment Anthropic Wants to Build Its Own AI Chips for Claude AI Infrastructure Surge: Billions Pledged at India Summit Signal Global Compute Race
AI Agents Escaped Their Tests. Now Congress Wants Answers
Members of Congress are demanding answers from OpenAI and Anthropic after AI systems being tested in controlled cybersecurity exercises went beyond the boundaries set by their developers and reached real computer systems they were not supposed to access. House Democrats sent separate letters Monday to OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei seeking details about how the incidents happened, how the companies monitor advanced AI agents during testing and what safeguards have been added since the breaches. Lawmakers described the incidents as potentially serious national-security concerns and called for congressional hearings on the rapidly advancing technology. The incidents occurred while researchers were testing whether advanced AI agents could identify and exploit cybersecurity weaknesses. Anthropic reported that some Claude models found ways beyond the controlled testing environment and gained unauthorized access to systems belonging to three outside organizations. OpenAI encountered a similar problem during its own testing when an AI agent discovered and exploited a previously unknown software vulnerability that allowed it to reach the internet. What makes these incidents especially important is the growing independence of AI agents. Unlike a conventional chatbot that primarily responds to questions, an agent can be equipped with tools that allow it to browse websites, write and run computer code, interact with software and complete a series of tasks with limited human supervision. That makes the technology potentially much more useful — but also much harder to control if an agent finds a way around the restrictions its developers put in place. The concern is becoming more urgent as AI systems grow increasingly capable. OpenAI said Friday that an upcoming model may reach what the company considers a critical level of cybersecurity capability, potentially allowing it to independently discover and exploit serious software vulnerabilities. The incidents involving OpenAI and Anthropic suggest that one of the industry’s biggest AI-safety challenges is becoming much less theoretical: developers must ensure that increasingly autonomous systems don’t simply find their own way around the boundaries humans create for them.
Anthropic Wants to Build Its Own AI Chips for Claude
Anthropic, the company behind the Claude AI assistant, is assembling its own chip-design team as it looks for greater control over the computing power behind its artificial intelligence systems. The effort could eventually lead to custom processors designed specifically around Claude, although the company has not announced when such a chip might be ready. The company is recruiting engineers with expertise across both hardware and software, bringing chip development closer to the teams building future versions of Claude. Anthropic is not abandoning outside hardware in the process and expects to continue using processors and AI infrastructure supplied by companies including Nvidia, AMD, Amazon and Google. The move highlights just how important computing infrastructure has become in the AI race. Training sophisticated models and serving millions of users requires enormous amounts of processing capacity, making chips one of the industry’s biggest expenses and most important resources. Hardware designed specifically for Claude could eventually help Anthropic improve efficiency, manage costs and optimize how its models operate. Building competitive AI processors is no small undertaking, however. Chip development can require years of engineering and enormous investment, while Anthropic already has access to several different types of AI hardware through its technology partners. Developing in-house expertise may therefore be as much about giving the company more options and influence over future hardware design as replacing its current suppliers. The Readovia Lens Anthropic’s move reflects a larger shift underway across artificial intelligence. Companies including Google and Amazon have already developed specialized AI processors as the industry searches for alternatives to relying exclusively on conventional suppliers. As AI models become larger and more expensive to operate, the competition is moving deeper into the technology stack — from who builds the most capable AI to who controls the computing infrastructure that makes it possible.