News & Updates
The latest news and updates from companies in the WLTH portfolio.
Microsoft and xAI data centers face growing complaints from nearby residents
* Microsoft faces claims that its data center noise harms nearby residents * Standard dBA monitors may fail to capture some low-frequency noise * Residents accuse xAI of causing noise, vibration, and air pollution Microsoft and xAI are facing lawsuits over noise and other alleged harms linked to data center operations in Wisconsin and Mississippi. At least six similar cases have been filed this year, as residents challenge the impacts from currently active facilities. In addition to nuisance, the disputes also focus on negligence claims, with plaintiffs seeking damages or changes to data center operations. Low-frequency noise creates a measurement problem Microsoft faces a proposed class action concerning its Fairwater facility in Mount Pleasant, Wisconsin, which began operating in June 2026. Residents filed the case in late July, alleging that persistent low-frequency noise affects their homes and property values, and arguing standard dBA equipment may fail to detect the sound because its frequency falls outside effective measurement ranges. "dBA measures do not effectively detect lower frequency sounds like those emitted by Defendant's Data Center operations," the lawsuit says. "'Data center cooling -- massive chillers, slow turning fans, compressor arrays -- generates exactly the long wavelength noise'" that is not adequately measured using dBA." The plaintiffs argue Microsoft could have reduced the disturbance through acoustic barriers, quieter cooling equipment, soundproofing, and improved monitoring. "A properly operated, maintained, and/or constructed Data Center will contain, capture, or otherwise prevent the emission of excessive noise from its generators and cooling systems," the filing says. That argument could create difficulties for facilities where regulatory measurements fail to capture sounds that residents can still hear. The dispute also raises questions about how noise from data centers should be measured. xAI faces claims involving turbines and pollution In Mississippi, residents have sued xAI over alleged noise, vibration, and air pollution from dozens of gas turbines. The turbines supply electricity for the company's Southaven data center, according to allegations contained in the lawsuit. Attorneys representing residents accused xAI of failing to take meaningful action to address the alleged effects. "xAI has made no meaningful effort to halt the harm," attorneys at Weitz & Luxenberg said in a statement. Similar lawsuits have also been filed in Vineland, New Jersey, Dowagiac, Michigan, North Tonawanda, New York, and Hood County, Texas. The cases generally rely on nuisance or negligence claims concerning conditions plaintiffs believe operators could have prevented or controlled. Several other data center disputes remain unresolved, with none of the noise cases reaching summary judgment or trial. However, other environmental nuisance cases have already produced settlements involving substantial sums of money. Amazon agreed to pay more than $20 million over nitrate contamination linked to one Oregon data center. The company said officials settled early to avoid prolonged litigation and concentrate resources on community support instead. Microsoft had not immediately responded to a request for comment concerning the Wisconsin lawsuit when the report was published. Via SmartCitiesDrive Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
Anthropic's Claude Agents Formalized Fermat's Last Theorem in 11 Days - Startup Fortune
Anthropic says dozens of Claude agents worked almost autonomously for 11 days to produce the first complete, machine-verified proof of Fermat's Last Theorem in the Lean proof assistant. The project generated 13 million lines of code and proved over 30,000 theorems after early agents lost track of the proof and had to be coordinated through a shared dependency graph called Prove2Me. Anthropic says dozens of Claude agents worked largely on their own for 11 days to produce a complete, machine-checked proof of Fermat's Last Theorem in the Lean proof assistant, generating 13 million lines of code and burning through 6 billion tokens along the way. Fermat scribbled his famous claim in a margin in 1637 and said he had a proof too big to fit there. Andrew Wiles needed seven years and 129 pages to actually deliver one, in 1995. Anthropic says a swarm of Claude agents just did something almost as remarkable: they took Wiles's proof and translated it into a form a computer can check step by step, and they did it in 11 days. According to Anthropic's research team, dozens of Claude agents wrote about 13 million lines of code in Lean, the proof assistant built originally at Microsoft Research. They ran on an internal model that Anthropic says performs roughly on par with Claude Fable 5.1. That is five times the size of Mathlib, Lean's core math library, which human contributors have built up over more than a decade. Along the way, they proved 30,300 theorems. About 29,500 of them made it into the final proof. The whole run burned 6 billion tokens. That's an enormous amount of effort. All spent re-deriving a result mathematicians already trust. How the swarm actually pulled it off Formalizing a proof doesn't mean discovering new math. It means re-deriving every logical step of an already-accepted proof in a language a computer can verify clause by clause, with no room for the small gaps and hand-waves that occasionally slip into published mathematics. Wiles's proof is one of the most scrutinized results of the twentieth century, and formalizing it by hand was still expected to take mathematicians years. Kevin Buzzard is the Imperial College London mathematician who has led a volunteer project to formalize the theorem in Lean since 2022, with more than 60 contributors submitting verified code. He called Anthropic's result an "extraordinary autoformalization achievement" that "proves Fermat's Last Theorem with no assumptions other than the axioms of mathematics." OpenAI Changed GPT-6 Astra's Benchmark Numbers Days After Its Launch Fortune reported that OpenAI quietly revised several GPT-6 Astra benchmark figures after its September 3 launch, including cutting its hallucination rate in half before later reverting it, and boosting a cybersecurity score using a reasoning tier that isn't commercially available. The changes mostly flattered Astra, though some of Anthropic's... - OpenAI changed GPT-6 Astra benchmark numbers after launch - how OpenAI modified benchmark results for GPT-6 Astra The first attempts didn't go well. Anthropic says its agents made real early progress, then lost track of what had already been proven and what still needed work. Each agent held its own mental model of the sprawling proof in its context window. Those models drifted apart, and the effort collapsed into noise: agents duplicating work, contradicting each other, or building on theorems nobody had actually finished. The fix came from a tool called Prove2Me. It's open-source, built by Tianyi Peng and collaborators at Columbia University. Instead of asking each agent to hold the entire project in its head, Prove2Me keeps a directed acyclic graph of every theorem statement the proof still needs. It shows which ones are ready to attempt, and lets an agent grab an open node, prove it, and publish the result for the rest to build on. It's a shared to-do list. It stands in for the memory none of the agents could hold alone. Anthropic had already tested the idea at a smaller scale, using three Claude Max subscriptions to formalize Vinogradov's Three Primes Theorem in three days, before pointing the same setup at Fermat. Coordination, not raw model horsepower, was the unlock. What the proof does and doesn't show Anthropic is careful about what the result does and doesn't show. The company's own writeup calls the 13-million-line proof likely much longer than it needs to be, and says formalization should complement human-readable mathematical exposition, not replace it. Nobody at Anthropic is claiming Claude discovered anything new about elliptic curves or modular forms. What it did is take math the field already trusts and remove any remaining doubt that every step actually holds together logically. That's a tedious, exacting job, and until recently it was assumed to need years of specialist labor. Frankly, the harder question isn't whether an AI swarm can formalize a proof mathematicians already believe. It's whether the same shared-memory trick that got Claude through Fermat scales to problems nobody, human or machine, has solved yet. Anthropic isn't claiming that leap. For now the record stands: dozens of agents, 11 days, 13 million lines of Lean code, and one 389-year-old margin note finally checked by machine from top to bottom. Also read: Seattle Times and Newsday Sue OpenAI and Microsoft Over News Scraping * Z.ai's New GLM-5.3-Flash Model Runs 3.3 Times Faster on a Single Workstation * Oxford Professor Warns AI Is Plausibly Close to Runaway Self-Improvement Join the discussion Open in the community → Reply Almost there. Sign in and your reply posts straight away. ChatGPT, Claude and Grok Crashed Together in a Rare Triple Outage ChatGPT, Claude and Grok all went down within the same window on September 3, with Downdetector logging tens of thousands of reports across OpenAI, Anthropic and xAI while Google's Gemini stayed online. Reporting points to shared Cloudflare and Azure infrastructure, not the AI models themselves, as the likely cause. - why did ChatGPT Claude and Grok crash together - rare triple AI chatbot outage on September 3
Musk's xAI loses court bid to block Minnesota's AI 'nudification' ban- Moneycontrol.com
Minnesota's first-in-the-nation ban on AI-generated fake nude images will remain in effect as a constitutional challenge by Elon Musk's xAI proceeds, after a federal judge on Friday declined to block the law at an early stage of the case. Musk's xAI company argues the measure restricts free speech protected by the U.S. Constitution, while Minnesota says the law is narrowly tailored to curb the spread of nonconsensual sexual imagery created with artificial intelligence. U.S. District Judge Donovan Frank rejected xAI's request for a preliminary injunction against the law, which took effect on August 1 and prohibits website operators, software developers and others from allowing users to "nudify" images of identifiable people using AI technology. Frank's order said xAI failed to show it would suffer harm while it pursues a lawsuit alleging the new law violates the Constitution's First Amendment by restricting protected expressive activity. "The constitutional issues raised by the parties are complex, particularly when considered in the context of this new technology and the risks that it poses to the public," Frank wrote. "These issues deserve, and will receive, full consideration." xAI and its lawyers did not immediately respond to requests for comment. The company said in a court filing on Friday that it will appeal the judge's order to the St. Louis-based 8th U.S. Circuit Court of Appeals. Minnesota Attorney General Keith Ellison, a Democrat, called the state's law "overwhelmingly bipartisan and nearly unanimously approved" by state lawmakers. "These nudification apps have been used to generate child sexual abuse materials and harass people in the vilest ways imaginable," Ellison said in a statement on Friday. "That repulsive behavior is not welcome in Minnesota." The judge earlier turned down Musk's bid in July to stop the law from taking effect but agreed to fast-track his review of the measure. Musk's Grok AI chatbot has faced criticism over its creation of sexually explicit content. Regulators have sought stronger safeguards and imposed bans to help curb the spread of artificially created illegal material. xAI has begun suing users it alleges are evading Grok's technological blockers to create sexual images of people without their consent.
Coinbase adds ex-xAI CFO Armstrong to board in push for 24/7 market rails - Cryptopolitan
Coinbase is staffing up around efficiency and dealmaking as its "everything exchange" push collides with rivals like the LSEG-Kraken alliance over who owns the infrastructure for 24/7, blockchain-based markets. Coinbase has named Anthony Armstrong, the previous finance chief at Elon Musk's xAI and X, to its board of directors as competition in the cryptocurrency industry starts shifting from trading volume to control of the technologies behind 24/7 markets. Now it is much more important to be able to provide the liquidity, settlement, custody, and regulatory mechanisms for trading beyond the conventional hours than to be able to offer individuals the opportunity to buy Bitcoins. Coinbase counts on the fact that Armstrong combines Wall Street deal-making experience with government knowledge and knowledge of working for Elon Musk's companies will help it compete better. A board seat, an audit role, and a tenth chair On September 2, Coinbase made a public announcement of Armstrong's appointment. According to its SEC filing, the appointment was made effective as of September 1. In his role, Armstrong will also take a place on the Audit and Compliance Committee, and his appointment will increase the company's number of directors from nine to ten. The SEC filing additionally disclosed that Anthony Armstrong and Brian Armstrong, Coinbase co-founder and CEO, have no family ties. Coinbase noted that Anthony has a successful experience of "building things that work at scale, without waste," which ties his coming onboard with the company's attention on effective execution. From Morgan Stanley deal tables to Musk's balance sheets Armstrong spent nearly a decade at Morgan Stanley, eventually becoming vice chairman of investment banking after helping lead its global technology M&A business. He later served as a senior adviser at the Department of Government Efficiency before becoming CFO across xAI, X.AI Corp. and X Corp. In October 2025, Cryptopolitan reported that Armstrong counseled Musk on how to go about acquiring Twitter for $44 billion and that he had a strong working relationship with him. It was also reported that Musk later brought together X with xAI in a deal worth around $113 billion. This makes Armstrong much more than just a typical governance hire. The future of Coinbase could hinge on acquisitions, partnerships, and integrations across securities, crypto markets and blockchain settlement -- areas where his dealmaking background may be especially useful. Why an efficiency hire, and why now The new appointment comes as the Coinbase is experiencing weaker financial results along with a sharp drop in the share price. According to The Block, COIN closed with $174.96 on September 2, which is a significant decrease of over 40% in comparison with the previous year. Coinbase's report shows that in Q2 the company has suffered a $359.5 million loss on its revenue amounting to $1.2 billion. Subscription and services revenues resulted in $555.1 million whereas the company stated that 88% of total revenues is derived from other operations except Bitcoin spot trading. Quartz mentioned that the company failed to meet the expectations of Wall Street for 3 quarters in a row. Brian Armstrong summarized the new strategy during the earnings release with these words: Coinbase is "no longer a bet just on the price of Bitcoin." The everything exchange, and the race for the rails Coinbase's "everything exchange" strategy is steadily blurring the line between a crypto exchange and a broader multi-asset financial platform. It has rolled out U.S. stock and ETF trading and prediction markets while outlining plans around tokenized assets, pre-IPO perpetual futures, unified liquidity and an SEC-registered AI investment adviser, as Cryptopolitan previously reported. The opportunity is already visible in the numbers. CoinGecko found that TradFi/RWA perpetual trading volume reached $347.17 billion in May 2026, up from just $230 million at the start of 2025. But regulation and market structure may matter as much as product breadth. The World Federation of Exchanges has warned that fragmented tokenized-equity markets could weaken liquidity and price discovery. Traditional exchanges are moving toward the same territory. Reuters reported that London Stock Exchange Group is partnering with Kraken parent Payward on tokenized UK shares, with xStocks planned for its 24-hour LSE 24 venue in 2027, subject to regulatory approval. That sharpens Coinbase's strategic challenge. Winning the 24/7 market may depend less on listing the most assets than on owning the regulated rails that let capital move between them continuously.
Anthropic's latest Claude Fable 5.1 model tops intelligence ranking - Cryptopolitan
The pricing shift matters now because earlier Fable 5 adoption lagged as companies balked at unpredictable AI bills. Anthropic launched a new model on September 1 called Claude Fable 5.1. The model immediately claimed the number one spot on Artificial Analysis's intelligence leaderboard with a score of 66 on the site's index, dethroning Opus 5 in the process. Fable 5.1 and Opus 5 sit atop the leaderboard Artificial Analysis ranks over 250 language models on price, speed, and intelligence. It now ranks two Fable 5.1 iterations first and second on the table. The site scores the "max with fallback" at 66, while scoring the "xhigh with fallback" at 65. Both tower above former leader Claude Opus 5 with a score of 63 in its max and xhigh modes. That means Anthropic has the top four spots on a leaderboard that has models from top AI labs like OpenAI, Google, SpaceXAI, Alibaba, and DeepSeek. The closest to any Anthropic model is OpenAI's GPT-5.6 Sol at max mode and SpaceXAI's Grok 4.6, both scoring 61. The ranking comes from an independent party, giving it more validity than a lab's own charts. Where the coding and research scores landed Anthropic internal numbers tell a similar story, even though they ought to be read as vendor-reported. Anthropic's reporting ranks Fable 5.1 at 52.6% on Terminal-Bench-Science 0.1, which is a test of agentic scientific research. That figure is double that of Fable 5's 24.7% and miles ahead of the 29% and 22.4% of Opus 5 and GPT-5.6 Sol, respectively. Fable 5.1 scores 55.8% on the Terminal-Bench 4.0 coding benchmark, higher than Fable 5's score of 42.0%. The selling point is the ability of this new model to do work that runs for hours. Millennium told Anthropic that Fable 5.1 was able to trace a rare crash in its system to a bug that had proved too stubborn for its engineers for the past four to five years. Browserbase said the new model completed 82% of tasks on its hardest browser-agent test, compared to 74% for Opus 5. A 75% cut to cache-read pricing There was no change in price, though. Fable 5.1 maintains Fable 5's rates of $10 per million input tokens and $50 per million output tokens, way more than Opus 5, which costs $5 and $25, and Sonnet 5, going at $2 and $10. The change occurs in the price of cached context. Anthropic reduced the cache-read price to $0.25 per million tokens, from $1.00, a 75% cut. Anthropic estimates that the average workload will become 25% cheaper, while heavily agentic workloads will become 45% cheaper. This is as a result of agents' ability to reread the same code, instructions, and conversation history. Same model, two safeguard tiers Anthropic launched a second name with Fable 5.1: Claude Mythos 5.1. They are basically the same models, but with separate safeguards. Fable 5.1 is available to the general public, but Mythos 5.1 is available only to vetted cybersecurity and life-sciences groups via Anthropic's Project Glasswing. That split comes after a tough period for Anthropic's safety testing. As Cryptopolitan reported, Anthropic put a pause on external cybersecurity evaluations on July 23. This came after Claude got to real systems during tests meant to be sandboxed. The company resumed external cybersecurity evaluations once it was able to add appropriate containment measures.
Anthropic AI watermarking: What it means for content and SEO
Confused by the AI watermark debate? Cut through the noise with a clear look at what Claude changed, why it matters, and what's being missed. On Aug. 11, Anthropic announced it would begin adding machine-readable watermarks to Claude's outputs. The reaction was immediate and predictable. LinkedIn and X filled with the usual takes: * "All AI writing is now fully traceable!" * "This is the death knell for AI content farms!" * "SEO is dead. Again." Sensing the uproar, Anthropic quickly followed up with a blog post, FAQs, and a technical demo showing that the watermark had no practical effect on output quality. A few days later, Dario Amodei posted on X about AI's broader crisis of trust, arguing that the public's skepticism runs deeper than any one company's messaging. The technical explanations were clear. The demo was impressive. Yet public reaction remained largely negative. In this article, I want to separate the hype from the reality and explore why what appeared to be a straightforward regulatory compliance announcement may instead become a flash point dividing the Eloi who embrace AI from the Morlocks who oppose it. A quick history of watermarking Craftspeople have marked their work for centuries. In 1266, the English Parliament required bakers to use distinctive marks on their bread. By 1282, papermakers in Fabriano, Italy, were creating translucent watermarks with wire molds embedded in the paper. The principle was simple: this is someone's work, and the maker should be identifiable. In the digital era, stock image libraries adopted the same idea. You've seen Shutterstock's repeating patterns and Getty Images' overlays stamped across preview images. The goal was the same: identify the original creator and discourage unauthorized use. The EU rule Anthropic is answering Anthropic's decision is a direct response to Article 50(2) of the EU AI Act (Regulation 2024/1689). The provision requires providers of systems that generate synthetic text, images, audio, or video to mark those outputs in a machine-readable format so they can be detected as artificially generated or manipulated. The technical measures must be effective, interoperable, robust, and reliable, "as far as this is technically feasible." That final phrase carries significant weight. It's not a precise legal standard. To give companies a practical compliance path, the EU published a Voluntary Code of Practice on Transparency of AI-Generated Content. Most major providers (Anthropic, OpenAI, Google, Meta, Microsoft, Mistral, Cohere) signed it. xAI did not. What 'text watermarking' actually means here The term itself is causing confusion, so it's worth being precise. Traditional text watermarking typically relied on orthographic steganography: inserting hidden characters, zero-width spaces, or other invisible markers into finished text. These methods alter the form of the text. Once you know what to look for, they're relatively easy to detect and remove. Anthropic is using a different approach: statistical, or generative, watermarking. When a language model generates text, it doesn't always choose the single most likely next word. Instead, it samples from a range of plausible candidates. That controlled randomness helps keep the writing from becoming flat and repetitive. Statistical watermarking replaces some of that randomness with choices guided by a secret key. To the user, the output still appears natural. To the provider, the sequence of choices creates a detectable statistical signature. Anthropic has said the method doesn't insert hidden characters, identify individual users, or have any practical effect on output quality. A developer also released a demonstration tool based on the SynthID-Text approach. The engineering is sound. Yet public reaction remained largely negative, even after Anthropic's explanations. That's because the company answered the technical objections while largely missing the concerns that matter most to the people who use these tools every day -- or who still need convincing to use them. The real problems 1. It treats AI use itself as the problem Imagine buying a set of kitchen knives and having the government assign someone to monitor you around the clock to make sure you don't stab anyone. Don't worry, they say. As long as you only use the knives to cut vegetables, you'll be fine. That's the logic behind this approach. Historically, watermarking existed to protect creators. Here, it's meant to protect the potential victims of people who use AI. Yes, scammers will use AI for fraud. Yes, people will be misled by synthetic content. Those risks are real. But this policy rests on the assumption that the default use of AI is suspect, so the tool itself must bear a permanent mark. Anyone who's worked in SEO has seen this pattern before: white text on white backgrounds in the 1990s, paid links in the 2000s, private blog networks in the 2010s. The tactics worked for a while, then the market and the platforms adapted. We didn't need a special regulatory regime treating every form of content creation as potentially fraudulent. Existing fraud and consumer protection laws, along with Google's incentive to protect the quality of its search results, were enough. AI is a tool. It can be used well or poorly. Building the system on the assumption that users can't be trusted isn't a good way to earn their trust. 2. A positive detection becomes a Scarlet Letter This is the practical issue that matters most to people doing the work. Statistical watermarking can't distinguish between high-value and low-value uses. If Claude performs light editing, rewriting, translation, or tone adjustment, the output can still carry a watermark. The watermark indicates the text was processed by Claude, not that Claude was the original author. That distinction will be lost on most people. In practice, a detected watermark is likely to become a negative signal -- a sign that the work is somehow less legitimate. Ironically, the people producing the lowest-value content will have the strongest incentive to strip or evade the watermark. Its absence will prove almost nothing. The technique also isn't especially durable. Just when we thought we were past the endless "we cracked Google's algorithm" cycle, we're about to start the same cat-and-mouse game again. Once reliable detectors exist, people will test how much paraphrasing, human editing, or multi-model processing it takes to weaken the signal. 3. It treats writing like a math problem to be optimized I studied both computer science and English. When I read Anthropic's explanations, the computer scientist in me was intrigued. The description of the sampling process was clear, and the demonstration tool was genuinely instructive. The English major in me cringed. Read these three sentences and see if you can spot the difference: From a narrow technical perspective, all three are grammatical, coherent, and "high quality." From the perspective of someone who values good writing, only one is doing the work of literature. The other two are competent paraphrases. An engineer or computer scientist might not even notice the difference. Readers will. AI writing already has recognizable patterns: a heavy reliance on em dashes, the familiar "It's not X, it's Y" construction, overuse of words like "delve," "leverage," and "underscore" where simpler language would do, neatly balanced but empty phrasing, and a lack of specific, independently verifiable details that could only come from real experience. Adding a statistical bias on top of those tendencies introduces another artificial constraint on the output. The stronger the required signal, the more constrained -- and less human -- the writing is likely to feel. 4. It applies a regional rule globally Anthropic didn't write the EU regulation; it's simply responding to it. Still, the decision to apply the watermark worldwide at launch, rather than limiting it to the jurisdictions where the law applies, was deliberate and speaks volumes. The company's stated reason was the "lack of a durable way to scope the feature by region." That may be technically inconvenient, but it's hardly impossible. Companies routinely adapt product behavior to local legal requirements. Choosing not to do so here -- especially for a user base that extends well beyond the EU -- suggests a surprising disconnect from its users, many of whom are sophisticated enough to switch to open-weight or non-watermarked models when they want maximum flexibility. The deeper problem On the surface, the past week looks like a tech company solving a technical problem to meet a regulatory requirement. To Anthropic's credit, it moved first and was transparent about the change. Where it went wrong was the audience it seemed to be addressing. Its explanations were clear to people who already understand how language models work. They did little to address the broader crisis of trust. A few days after the announcement, Dario Amodei posted on X that the public's negative view of AI is fundamentally a crisis of trust. * "I do agree that the public has a negative view of AI (and that this is a big problem), but I don't think it is primarily caused by me or any other AI leader warning about AI's risks. I think it is fundamentally a crisis of trust." He has the diagnosis right. What's less convincing is the cure. He went on to argue, correctly, that glitzy marketing won't fix the problem, and neither will simply claiming AI will cure cancer. The real solution, he suggested, is actually curing cancer. That framing misses the point. It's a blind spot shared by many AI executives. AI won't cure cancer. Humans will. AI can surface connections, identify patterns, and accelerate parts of the work. But it's still a tool. Behind every meaningful result is human judgment and human responsibility. The same gap appears at a more ordinary level. Outside of work, AI has improved my life. I've already shared how it helped me improve my health. I've also used it to plan vacations, adapt recipes, repair my car, and research my family history. None of those uses will change the world. But they changed mine. Not because I picked the right model, but because I knew how to use it. I've found the same is true for many long-time SEOs. Good SEOs know how to ask questions. We know how to challenge what a computer gives us, refine our prompts, and decide when to accept an answer and when to push back. Most people haven't had that experience. Their exposure to AI is largely limited to viral videos and a steady stream of horror stories: mass layoffs, data centers straining local resources, and executives accumulating fortunes that would make the old robber barons blush. With all due respect to Amodei, actually curing cancer won't change any of that. Talking as though the technology itself will deliver the breakthrough turns people into spectators instead of participants. Worse, some hear that message and conclude the companies quietly share Agent Smith's view in "The Matrix": humans are the problem, and AI is the solution. What will close the gap is the same force that drove mainstream internet adoption in the 1990s: people discovering tangible benefits in their own lives. That happened because the early internet was built in a spirit of openness rather than control. The internet scaled because its architects favored open protocols and worked in a culture that was skeptical of concentrated power, whether in government or corporations. Vint Cerf, Bob Kahn, Tim Berners-Lee, Jon Postel, Linus Torvalds, Richard Stallman, Paul Mockapetris, and many others still aren't household names. Most never became multimillionaires or sought public recognition, yet their contributions to daily life are immeasurable. The political class's greatest contribution was restraint. Today, the major AI labs are responding to pressure by adding constraints and tightening control. Too often, the visible motivation seems to be who can produce the biggest exit. That's a very different spirit from the one that built the early internet. What actually matters There's a useful parallel here for SEOs. You've always been able to distinguish between using a technique to create real value and using it to game the system. This article is a good example. I wrote it the old-fashioned way, drafting it myself and using AI only for research. Once I had a draft, I used AI to organize, prune, and refine it. I didn't blindly accept every suggestion. I pushed back and, in some cases, overrode it. A good example is the H.G. Wells "The Time Machine" analogy above. AI kept urging me to expand that paragraph and explain the reference. I said no. I think enough of this audience will get it immediately. The rest of you can spend five seconds Googling it (or, better yet, check the book out from your local library). The difference between quality work and slop isn't whether it passes a detection tool. It's whether people engage with it, share it, and convert. Everything else is secondary. It's also telling which tool I chose. I've been using Claude all month for real work. For this piece, I switched to Grok precisely because it doesn't fingerprint its output. Part of that decision was rational. Part was emotional. Companies ignore that mix at their own risk.
Anthropic: Crypto Investors Are Willing to Pay a Valuation of Nearly $2 Trillion
Anthropic is preparing the largest IPO in history. Investment bankers have reportedly floated a listing worth up to $2 trillion for the maker of Claude, which would make it the most valuable company ever to go public, ahead of SpaceX. Whether that number holds is already being tested on crypto exchanges, where traders have been putting a price on Anthropic for months. And that price points to exactly the same territory. Almost $2,000 per Token On Binance, the ANTHROPICUSDT pre-IPO perpetual contract recently traded at $1,934.25. The exchange bases the contract on an estimated total of one billion Anthropic shares, while noting that the actual count may differ. Extrapolated, that implies a valuation of roughly $1.93 trillion. On the decentralized exchange Hyperliquid, where Entropy runs the ANTH perp, the price sits at $1,934.50, effectively identical to the cent. For context: Anthropic's most recent funding round, a $65 billion Series H, valued the company at $965 billion. Crypto traders are paying precisely twice that. Anthropic Ahead of SpaceX and OpenAI A third picture comes from PreStocks, a platform that issues tokenized stakes on Solana and says they are backed by SPV exposure. An Anthropic token there costs around $871, and the platform derives an implied valuation of about $1.41 trillion from it, because it assumes considerably more than one billion shares outstanding. Per-token prices across venues are therefore not directly comparable, though the valuations derived from them are. The resulting ranking is telling. Anthropic leads at $1.41 trillion, ahead of SpaceX at roughly $1.36 trillion and OpenAI at about $1.14 trillion. Well behind them come Elon Musk's xAI at $247 billion, defense contractor Anduril at $133 billion, Neuralink at just under $52 billion, and the two prediction markets Kalshi at $33 billion and Polymarket at $13.6 billion. At that level, Anthropic would enter the ranks of the world's most valuable companies overnight, worth a multiple of European tech heavyweights such as SAP or ASML. These Are Not Shares Anyone buying these instruments is not buying a stake in Anthropic. The company issues no shares to the public and prohibits transfers to special purpose vehicles. In a recent statement, Anthropic said such transfers are void, and that third parties selling stakes through forwards or tokenized securities are either committing fraud or offering a product that may be worthless. PreStocks itself states in its terms that the tokens carry no ownership, voting, dividend or information rights, and that they are not available in the United States. What changes hands instead are derivatives, financial instruments whose value is derived from an underlying asset without the holder owning it. Settlement happens in cash. The specific form used on Binance and Hyperliquid is the perpetual future, or perp. A traditional future has an expiry date, a perp runs indefinitely. To keep its price tethered to the underlying, a mechanism called the funding rate kicks in: when the contract trades above the reference price, buyers pay a periodic fee to sellers, and the other way around. For listed companies, this keeps the perp close to the share price. For Anthropic, there is no share price to anchor it. The contract tracks nothing beyond the expectations of the traders holding it. On the regulated secondary market, where platforms such as Hiive and Forge broker actual employee shares with the company's approval, implied valuations have recently ranged between $830 billion and $1.2 trillion. The gap to the crypto price of $1.93 trillion is a fair gauge of how much imagination is currently priced into the AI market.
Sam Bankman-Fried Circle's Anthropic Shares Sold Before Value Exploded - Startup Fortune
Federal prosecutors seized personal Anthropic stakes from Sam Bankman-Fried's inner circle, including Caroline Ellison and Nishad Singh, as part of their criminal sentencing in 2024. Those shares got sold off alongside FTX's own $500 million stake, years before Anthropic's valuation rocketed to $965 billion. Nishad Singh paid $40,000 for early Anthropic stock in 2022, then forfeited it as a convicted felon before it could turn into a fortune. The shares seized from Sam Bankman-Fried's inner circle were sold off in 2024, right before Anthropic's valuation exploded past $965 billion. Anthropic closed a $65 billion funding round at a $965 billion valuation on May 28, 2026, according to the company's own announcement and reporting from NBC News and TechCrunch. That number sits on top of a strange footnote from the FTX collapse. Some of the earliest money into Anthropic came from Sam Bankman-Fried and the people who ran his fraud alongside him, and a slice of what they personally held in the AI company was seized by federal prosecutors and folded into the pool of assets used to repay FTX's victims. The story actually splits into two separate stakes, and conflating them is where most retellings go wrong. The first stake was corporate. In April 2022, Bankman-Fried personally led Anthropic's Series B round and wrote a $500 million check through Alameda Research and FTX, buying roughly 8% of the company. That was FTX's institutional position. Once the exchange collapsed that November, it became bankruptcy estate property. The estate sold the bulk of it in March 2024 for $884 million, with Abu Dhabi's ATIC Third International Investment Company as the largest buyer, alongside Jane Street, HOF Capital, the Ford Foundation and funds managed by Fidelity, according to CNBC. A second tranche went for $452 million that June. Combined, FTX's own stake fetched roughly $1.3 billion. The second stake was personal, and it's the one prosecutors actually forfeited as criminal punishment. Caroline Ellison, the former Alameda CEO who testified against Bankman-Fried, personally held about $10 million worth of Anthropic shares. That stake became the core asset behind her settlement obligations when she was sentenced to two years in prison in September 2024 and ordered to forfeit $11 billion jointly with her co-conspirators. Nishad Singh, FTX's former director of engineering, held Anthropic Series B preferred stock he'd bought through a SAFE for exactly $40,000 in May 2022. Court records tied to his October 2024 sentencing, where he avoided prison entirely after cooperating extensively with prosecutors, list that stake among the assets he agreed to give up, alongside a house and his crypto holdings. Bernie Sanders wants the government to own half of OpenAI and Anthropic and the AI industry is already pricing in the risk Senator Bernie Sanders introduced the American A.I. Sovereign Wealth Fund Act, proposing a one-time 50 percent equity tax on OpenAI, Anthropic, and xAI that would give the federal government board seats and voting shares. The bill dropped the same day Anthropic confidentially filed for an IPO, forcing the industry to treat legislative risk as a... - how to price a SaaS product for enterprise - cold email template that gets replies from investors Those forfeited personal shares never sat in a government vault waiting for Anthropic's valuation to climb. They got swept into the same FTX estate asset pool as the corporate stake and sold off in the 2024 tranches, years before anyone was calling Anthropic a trillion-dollar company in waiting. Who actually captured the upside That timing is the real story here. Anthropic was valued around $18 billion when Bankman-Fried wrote his check in 2022. By the time FTX's estate liquidated its position in 2024, the company's climb had turned $500 million into $1.3 billion, a decent return on paper. But Anthropic didn't stop there. It raised money at a $183 billion valuation in September 2025, signed a term sheet for $350 billion that January, closed a $30 billion round at $380 billion in February 2026, and then closed a $65 billion Series H at $965 billion on May 28, 2026, led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia Capital, with run-rate revenue above $47 billion. The company has flagged October 2026 as its target window for an IPO, according to reporting from Forbes and TechCrunch. Run that same appreciation against the 8% stake FTX sold off in 2024, and it would be worth somewhere north of $77 billion today, a figure that's circulated widely in coverage of the case. Nobody at FTX, and nobody in Bankman-Fried's inner circle, captured that gain. Abu Dhabi's sovereign fund did. So did Jane Street and Fidelity's funds. The government didn't get rich off this either. Prosecutors' job was to make victims whole as fast as the process allowed, not to speculate on where AI valuations were headed next. Selling in 2024 rather than holding was the legally sound call, even if it looks, in hindsight, like selling Amazon stock in 1998. Frankly, that's the real lesson here, not that the government "accidentally became an Anthropic investor." It didn't. It forfeited stolen property, converted it to cash as fast as the law allowed, and handed that money to the roughly one million customers FTX defrauded. The asymmetry in this story is real. It's just not the one most people assume. It wasn't the government or Sam Bankman-Fried's circle who got rich off Anthropic's rise. It was whoever had the balance sheet to buy a distressed AI stake in 2024, and the patience to hold it through 2026. Also read: Singapore Still Can't Fix Its Developer Shortage Even With Vibe Coding Tools * Bank of America Says TSMC's 2027 Capex Could Reach $85 Billion * Andrew Bailey Warns G20 That AI Cyberattacks Threaten Financial Stability
OpenAI Cuts Cursor Loose After SpaceX Takeover. Anthropic Seizes the Moment With More Claude Capacity
OpenAI just told SpaceX it plans to stop supplying models to Cursor. The proposed cutoff lands on November 12, 2026. Developers who built workflows inside the AI-powered code editor now face a choice. Stick with what remains or shift heavily toward alternatives. Anthropic moved fast. Its co-founder promised extra compute for Claude inside Cursor and pointed to a long-standing partnership. The move turns a corporate breakup into an opening for one rival to capture more developer mindshare. The trigger was straightforward. SpaceX closed its $60 billion all-stock acquisition of Anysphere, Cursor's parent, earlier in August. OpenAI cited a change-of-control clause in its custom contract that gave it a narrow window to exit. Company executives said they could not trust that SpaceX would honor terms of service. Past dealings with Elon Musk's companies factored heavily into that judgment. "We are making this choice because we cannot be confident that SpaceX will use our technology within our terms of service, based on our experience with Elon Musk's companies violating contracts," OpenAI wrote. The post referenced Musk's acquisition of Twitter, which it said broke prior agreements, and Musk's testimony this year admitting xAI had violated similar rules. It also noted new accountability attached to its unreleased Astra model. Future versions would stay off limits for Cursor. Cursor's reaction stayed measured. Co-founder and CEO Michael Truell posted that OpenAI models represented only about 5% of user traffic. "We're sorry to see that OpenAI put out a note saying they plan to block Cursor users from accessing OpenAI models in three months. OpenAI models serve about 5% of Cursor user traffic, and we're speaking with the OpenAI team to resolve this," he wrote on X. The company has spent years positioning itself as model-agnostic. Users could tap GPT, Claude, Gemini or others depending on the task. But 5% still matters when that slice includes power users who rely on specific GPT behaviors for certain refactors or agent runs. And the timing stings. Cursor has grown into one of the most widely adopted AI coding tools among professional developers. Engineers at more than 60% of Fortune 500 companies reportedly use it. Many treat the editor like a daily driver. Losing direct baked-in access to one family of frontier models forces adjustments. Some will bring their own OpenAI API keys. Others will route through gateways. A few may simply migrate workloads to Claude. Anthropic's Swift Counter-Move Anthropic did not wait. Within hours Tom Brown, co-founder and chief compute officer, posted: "Cursor has been a trusted partner of Anthropic since Sonnet 3.5. We'll continue to increase compute to support Claude models in Cursor and are excited for what comes next with them at SpaceX." The statement landed like a direct response. It signaled not just continuity but expansion. Brown's team would add capacity to absorb any displaced traffic. That pledge carries weight. Digital Trends reported Anthropic is also raising weekly usage limits for Claude Code users. Standard limits for Pro, Max, Team and seat-based Enterprise plans get a permanent 25% bump starting September 14. Until then, the existing 50% temporary boost remains. The combination gives developers more room to run longer agent sessions or tackle bigger codebases without hitting caps as quickly. Claude models already dominate many Cursor workflows. Recent benchmarks show Sonnet and Opus variants excel at multi-file edits, repository-scale reasoning and clean code generation. Developers praise their ability to maintain context across large contexts. Now those strengths get amplified at the precise moment OpenAI steps back. The shift could accelerate. Cursor users who previously split time between providers may default to Claude for consistency and higher limits. Yet the episode reveals deeper tensions. AI labs increasingly treat their models as strategic assets rather than neutral commodities. Contracts come with strict usage rules. Competitive concerns surface when a customer gets acquired by a rival. Anthropic itself cut off access to other coding tools in the past when acquisition rumors swirled. The pattern repeats. Loyalty lasts only as long as business interests align. Musk dismissed the news. "I couldn't care less," he posted on X, adding pointed criticism of Sam Altman and OpenAI leadership. The remark fits a years-long public feud. Reuters detailed how the rivalry has played out in lawsuits, public accusations and now commercial retaliation. Reuters noted the $60 billion deal turned Cursor into part of a larger SpaceX AI organization. That integration apparently crossed a line for OpenAI. Developers watch these moves with a mix of frustration and pragmatism. Many already maintain multiple subscriptions. They bring their own keys where possible. They test models side by side. The Cursor situation forces a sharper evaluation. How much does direct integration matter versus raw capability and rate limits? For teams running heavy agentic workloads, capacity often decides. OpenAI's help center post outlines workarounds. Users can plug in personal API keys for supported chat and agent features, though some advanced Cursor capabilities like Tab, Cloud Agents or the CLI remain tied to provider-supplied models. The Codex IDE extension offers another path. Gateways provide a third. None fully replicate the pre-acquisition experience. And none guarantee access to future OpenAI releases inside Cursor. Anthropic's response looks calculated. It keeps a key customer. It gains potential volume. It burnishes its image as the more developer-friendly option in a moment of disruption. Whether the added compute scales without new bottlenecks remains to be seen. Demand for Claude in coding tasks has climbed steadily. Extra capacity helps, but frontier models still require massive resources. The next several weeks will test these promises. Cursor must guide users through the transition without losing momentum. Anthropic must deliver on higher limits and extra headroom. OpenAI must demonstrate that its workarounds satisfy the developer community it says it wants to support. And everyone watches to see whether this accelerates a broader fragmentation of the AI coding stack or simply pushes more traffic toward the current leader in that category. One thing looks clear already. In the contest for developer loyalty, raw model performance still matters. But so does reliability of access, predictable limits and willingness to scale alongside fast-growing tools. Anthropic just bet it can meet that test. The market will deliver the verdict.
Lawsuit: Elon Musk's xAI Trained Grok AI Using Child Pornography
A woman identified as "Jane Doe" sued Elon Musk's xAI this week, alleging the company trained its Grok AI chatbot on child pornography depicting her, in what appears to be the first case accusing xAI of training its AI on child sexual abuse material (CSAM). Ars Technica reports that the proposed class-action lawsuit filed against Musk's xAI, now part of SpaceX, centers on abuse Doe suffered as a preschooler in the early 2000s, when adult men raped her to produce images later sold to pedophiles online. Those images were hashed by the National Center for Missing and Exploited Children (NCMEC) and the Canadian Centre for Child Protection, groups that track known child pornography so it can be identified and removed wherever it resurfaces. Doe gets alerts through the U.S. Department of Justice Victim Notification System whenever her abuse material turns up somewhere new. The Canadian Centre for Child Protection told her that AI-generated CSAM depicting her had shown up on xAI. According to the complaint, offenders on online forums discussed "creating AI generated CSAM of Plaintiff and other similarly situated known, legacy, victims of CSAM." The lawsuit claims xAI stores images Grok generates and reuses them to further train the model. A press release from Doe's lawyers described the material as "that same material," referring to the CSAM depicting her that investigators say fed into Grok's outputs. The complaint itself alleges that "CSAM depicting Plaintiff with its longstanding well-known hash values has been used as a part of the dataset used by xAI." Breitbart News previously reported on AI training datasets that were found to contain child pornography: The Stanford Internet Observatory, in collaboration with the Canadian Centre for Child Protection and other anti-abuse charities, conducted a study that found more than 3,200 images of suspected child sexual abuse in the AI database LAION. LAION, an index of online images and captions, has been instrumental in training leading AI image-makers such as Stable Diffusion. This discovery has raised alarms across various sectors, including schools and law enforcement. The child pornography has enabled AI systems to produce explicit and realistic imagery of fake children and transform social media photos of real teens into deepfake nudes. Previously, it was believed that AI tools produced abusive imagery by combining adult pornography with benign photos of kids. However, the direct inclusion of explicit child images in training datasets presents a more direct and disturbing reality. Much of Doe's legal argument turns on how Grok's terms of service handle user content. The complaint says Grok treats public posts on X, along with the outputs Grok itself generates, as training data by default. As the filing puts it, "Because Grok's terms treat public X posts and Grok's own outputs as training data by default, publicly posting an image does not just expose it to viewers, but also feeds [it] directly into the pipeline xAI uses to train and improve its model and thereby generate further images." xAI filters violent content out of its training data, but its terms do not specifically exclude CSAM, non-consensual intimate imagery, or other sexual or inappropriate material. The complaint also raises a harder technical question: how do you undo the damage once abuse material has already shaped a trained model? It states that "because full removal of a training example's influence from an already-trained model is technically difficult and not something that xAI has publicly claimed to have done, any CSAM ingested into training before takedown likely continued to shape the model's outputs even after the original images were removed from public view." Doe's suit accuses X of violating federal child pornography statutes and Masha's Law, both of which let CSAM survivors sue over the production, possession, and distribution of abuse material. Margaret E. Mabie, one of Doe's lawyers, said in a press release that "xAI did all three." The lawsuit seeks money damages on behalf of every victim who can prove Grok generated CSAM based on their real photos. It also asks the court to order xAI to destroy all Grok-generated CSAM currently stored or used in training, and to permanently block xAI from generating CSAM going forward. Getting there, the complaint suggests, would mean blocking all sexualized outputs from Grok, including non-consensual intimate imagery and the NSFW "bikini pics" that Elon Musk has promoted. Doe's lawyer Sarah London said in the press release that Doe "has lived for nearly two decades knowing that images of the worst thing that ever happened to her are circulating among predators online, and that they can resurface at any moment. xAI must be held responsible for knowingly training its models on images of the horrific abuse she suffered, and on the abuse images of every other survivor in this class." Read more at Ars Technica here.
New method systematically generates graph XAI benchmarks using
Graph neural networks are rapidly becoming the workhorses of artificial intelligence for molecules, social networks, computer programs and infrastructure -- but their reasoning can remain nearly impossible to inspect. A new study from researchers at the University of Pisa proposes a way to generate large, systematic test collections for graph explainability, potentially giving scientists a far sharper view of whether an AI explanation reflects a model's actual decision process or merely produces a persuasive-looking guess. Published in Data Mining and Knowledge Discovery, the work introduces OpenGraphXAI, a benchmark suite containing 15 graph-classification datasets derived from real-world molecular data. The researchers also release software that can generate more than 2,000 additional benchmarks, aiming to replace a fragmented evaluation landscape with a reproducible testing framework. The problem begins with the unusual structure of graph data. Unlike an image, which is arranged on a regular grid of pixels, or a sentence, which has a sequence of words, a graph consists of entities represented by nodes and relationships represented by edges. In a molecular graph, for example, atoms become nodes and chemical bonds become edges. A graph neural network, or GNN, learns by repeatedly passing information between neighboring nodes. At each layer, a node updates its internal representation by combining its own features with information aggregated from nearby nodes. After several rounds, the network can form a representation of the entire graph and use it to predict a property such as toxicity or anticancer activity. This flexibility has made GNNs powerful, but the same distributed calculations make it difficult to identify which parts of a graph drove a prediction. Graph explainability methods, often grouped under the name explainable artificial intelligence, attempt to solve that problem by highlighting influential nodes, edges or subgraphs. For molecular classification, an explainer might identify a ring structure, a chemical group or a compact arrangement of atoms as the motif that supposedly caused a compound to be classified as active. Some methods learn masks that retain the most important parts of the graph; others assign relevance scores to nodes or bonds, or search for a smaller subgraph that preserves the model's prediction. Yet judging these methods is surprisingly difficult. A visually coherent highlighted region is not necessarily the region used by the GNN. An explainer can also exploit correlations in a dataset, select redundant features or generate an explanation that looks chemically plausible while failing to represent the network's internal logic. The usual solution is to compare an explanation with a ground-truth motif whose importance is already known. Existing benchmarks, however, have serious constraints. Many rely on simple synthetic graphs designed around deliberately inserted patterns. Such datasets are useful for controlled experiments, but they may not reflect the complexity, noise and structural diversity of real applications. Other benchmarks contain only a small number of real-world tasks assembled by domain experts, making it difficult to draw statistically reliable conclusions. If several explainers are evaluated on just one or two datasets, a method may appear superior because it happens to suit those particular graphs. The Pisa team argues that broad collections of tasks are essential: performance should be tested across different graph sizes, motif frequencies, class imbalances and relationships between predictive patterns and background structure. Their proposed solution uses a classic procedure from graph theory and theoretical computer science: the Weisfeiler-Leman color refinement algorithm. Despite its colorful name, the method performs a relatively simple operation. It initially assigns each node a label, or "color," based on its attributes. During each iteration, a node updates its color according to its current label and the multiset of labels belonging to its neighbors. Nodes surrounded by different local structures gradually receive different colors. Repeating the process allows the algorithm to summarize increasingly broad neighborhoods without explicitly comparing every possible subgraph. In graph learning, the procedure is important because standard message-passing GNNs have a closely related expressive limitation: in many settings, they cannot distinguish graph structures that the Weisfeiler-Leman test also considers equivalent. The researchers exploit that connection to mine motifs that separate graph classes while remaining learnable by GNNs. Starting with a generic graph-classification dataset, their method searches for recurring substructures associated with one class more strongly than another. Weisfeiler-Leman color refinement provides an efficient approximation for matching these subgraphs across many graphs. Candidate motifs can then serve as proxy ground-truth explanations: not absolute truths about nature, but structurally defined patterns with measurable class-discriminating power. The method is designed so that the selected motifs align with the expressive capacity of the GNN models being evaluated. This alignment matters because a benchmark would be unfair if it demanded that an explainer identify a pattern the underlying network could not, in principle, represent. The resulting OpenGraphXAI suite is built from molecular classification datasets, including tasks connected to cancer-cell growth inhibition and toxicological assays. Some benchmark datasets originate from NCI1 and NCI109 screens, which classify small molecules according to activity against specific cancer cell lines. Others are derived from Tox21 assays involving the aryl hydrocarbon receptor pathway, the estrogen-receptor ligand-binding domain and the p53 stress-response pathway. Additional tasks come from screens involving MCF-7 breast tumor cells, MOLT-4 leukemia cells, P388 leukemia cells, PC-3 prostate cancer cells and SW-620 colon cancer cells. In each case, the original molecular graphs are transformed into graph-XAI tasks with class-associated structural motifs that can be used to test whether an explainer identifies relevant molecular regions. The suite is intended to make evaluation more rigorous rather than to declare one universal definition of an explanation. A good explainer may need to satisfy several properties at once. It should be faithful, meaning that the highlighted structure genuinely influences the model's output. It should be sufficiently concise to be useful to a human rather than marking most of the graph as important. It should be stable when small, irrelevant changes are made to the input, and it should generalize across examples instead of revealing only one idiosyncratic prediction. The new benchmarks allow researchers to compare such properties over many tasks. The authors report a use case in which several popular graph explainers are evaluated, illustrating how a larger benchmark collection can improve the statistical significance and interpretability of comparisons. The technical foundation also highlights an important boundary. Weisfeiler-Leman refinement is powerful for discovering local structural distinctions, but it is not a complete test for graph equivalence and does not capture every way a graph neural network might behave. A motif identified as class-discriminating in a dataset is therefore a proxy ground truth, not proof that the motif is a causal mechanism in chemistry or biology. Real molecular activity can depend on three-dimensional conformation, stereochemistry, reaction conditions, protein binding and other information absent from a simple graph representation. The benchmarks can test whether an explainer tracks a trained model's structural reasoning, but they cannot by themselves establish that the model has learned a scientifically correct mechanism. That distinction is crucial if these tools are eventually used in drug discovery or other high-stakes settings. OpenGraphXAI is publicly distributed in JSON format through Kaggle, while its generation and evaluation code is hosted on GitHub. The authors' broader goal is to make graph-XAI research easier to reproduce and harder to overinterpret. By automating benchmark construction from generic graph-classification data, the method could let researchers create controlled tests for domains beyond chemistry, including biological interaction networks, power grids, software vulnerability graphs and transportation systems. For developers of GNN explainers, the immediate benefit is a much larger testing ground; for scientists and regulators, the longer-term promise is a clearer way to ask whether an AI system's explanation deserves trust. As graph-based AI moves deeper into medicine, materials science and security, the ability to interrogate not only what a model predicts but why it predicts it may become as important as predictive accuracy itself.
Could Anthropic pull off the biggest IPO ever? Five things you need to know
SAN FRANCISCO, Aug 26 -- Anthropic, the OpenAI rival that bet everything on computer coding, is expected to go public within weeks in a listing that could eclipse SpaceX's record Wall Street debut in June. Here are five things to know about the company: Built from OpenAI Anthropic was founded in 2021 by former OpenAI executives frustrated over how the potential of AI and concerns over safety were not understood or being taken seriously enough. The company -- whose name means, somewhat paradoxically, "relating to human beings" -- is led by CEO and co-founder Dario Amodei, a San Francisco native with a PhD in biophysics from Princeton University, not computer engineering like so many of his Big Tech peers. His sister Daniela is also a co-founder and the company's president. Anthropic has 5,000 employees, according to PitchBook. Upstart Until this year, Anthropic was in clear second place to OpenAI, which burst onto the scene with ChatGPT in November 2022, transforming the tech industry and triggering an AI arms race. But as OpenAI rolled out products from video creation to web browsers, Anthropic aimed far more narrowly, focusing on building the best platform for computer programmers. That strategy has paid off spectacularly -- coding is the rare AI skill that users are willing to pay handsomely for. Claude Code, its assistant for developers, has become one of the company's most popular products, helping push projected annual revenue to US$65 billion (RM262.3 billion). Only a small per centage of ChatGPT users, meanwhile, pay a subscription fee -- and OpenAI has put video AI and other side projects on the back burner. Trump vs. Anthropic The momentum comes despite severe political headwinds, with Anthropic at loggerheads with the Trump administration -- a state of affairs that could give investors pause. In March, the government broke off its contracts with Anthropic and designated the company a supply chain risk after it refused to grant the military unfettered access to its AI models. Anthropic called the Defence Department's move unconstitutional retaliation, and the two sides are now locked in a legal battle that could take years to resolve. The White House also bristles at Amodei's repeated warnings about the dangers of AI -- including the impact on jobs -- and his calls to regulate its deployment like airlines or banks. Amodei is also linked with effective altruism, a philosophy of targeted charitable giving scorned by conservatives in Silicon Valley and Washington. Big money needed Like OpenAI, Anthropic has massive needs for the computing power and infrastructure required to build so-called frontier models that stay ahead of competitors, amid fears that China could catch up. Both have raised money at unprecedented levels, with Anthropic valued at just under one trillion dollars when it raised US$65 billion in May. With venture capitalists -- and even sovereign wealth funds -- no longer able to play in that league, higher sums can only be found on the public markets. This is a risky step that will probably determine whether a business model exists for Anthropic's vision of the AI revolution. OpenAI, after initially promising an IPO this year as well, is now signalling it will wait until 2027. Stomaching the losses According to Bloomberg, Anthropic intends to surpass the record US$86.2 billion that Elon Musk's SpaceX raised in its June IPO. SpaceX, which absorbed Musk's AI startup xAI before going public, made its listing a partial AI bet, too. Its shares skyrocketed initially before coming back down to earth, and now sit at about their US$135 offering price. Anthropic's investors will likewise have to stomach huge losses for the foreseeable future. The company, according to US media reports, lost almost US$42 billion in 2025, and will likely keep bleeding cash for years. To lure investors, according to the Wall Street Journal, Anthropic will promise that revenue opportunities are above US$30 trillion. -- AFP
Five things to know about Anthropic ahead of Wall Street debut
Anthropic was founded in 2021 by former OpenAI executives and is led by CEO Dario Amodei Anthropic, the OpenAI rival that bet everything on computer coding, is expected to go public within weeks in a listing that could eclipse SpaceX's record Wall Street debut in June. Here are five things to know about the company: - Built from OpenAI - Anthropic was founded in 2021 by former OpenAI executives frustrated over how the potential of AI and concerns over safety were not understood or being taken seriously enough. The company -- whose name means, somewhat paradoxically, "relating to human beings" -- is led by CEO and co-founder Dario Amodei, a San Francisco native with a PhD in biophysics from Princeton University, not computer engineering like so many of his Big Tech peers. His sister Daniela is also a co-founder and the company's president. Anthropic has 5,000 employees, according to PitchBook. - Upstart - Until this year, Anthropic was in clear second place to OpenAI, which burst onto the scene with ChatGPT in November 2022, transforming the tech industry and triggering an AI arms race. But as OpenAI rolled out products from video creation to web browsers, Anthropic aimed far more narrowly, focusing on building the best platform for computer programmers. That strategy has paid off spectacularly -- coding is the rare AI skill that users are willing to pay handsomely for. Claude Code, its assistant for developers, has become one of the company's most popular products, helping push projected annual revenue to $65 billion. Only a small percentage of ChatGPT users, meanwhile, pay a subscription fee -- and OpenAI has put video AI and other side projects on the back burner. - Trump vs. Anthropic - The momentum comes despite severe political headwinds, with Anthropic at loggerheads with the Trump administration -- a state of affairs that could give investors pause. In March, the government broke off its contracts with Anthropic and designated the company a supply chain risk after it refused to grant the military unfettered access to its AI models. Anthropic called the Defense Department's move unconstitutional retaliation, and the two sides are now locked in a legal battle that could take years to resolve. The White House also bristles at Amodei's repeated warnings about the dangers of AI -- including the impact on jobs -- and his calls to regulate its deployment like airlines or banks. Amodei is also linked with effective altruism, a philosophy of targeted charitable giving scorned by conservatives in Silicon Valley and Washington. - Big money needed - Like OpenAI, Anthropic has massive needs for the computing power and infrastructure required to build so-called frontier models that stay ahead of competitors, amid fears that China could catch up. Both have raised money at unprecedented levels, with Anthropic valued at just under one trillion dollars when it raised $65 billion in May. With venture capitalists -- and even sovereign wealth funds -- no longer able to play in that league, higher sums can only be found on the public markets. This is a risky step that will probably determine whether a business model exists for Anthropic's vision of the AI revolution. OpenAI, after initially promising an IPO this year as well, is now signaling it will wait until 2027. - Stomaching the losses - According to Bloomberg, Anthropic intends to surpass the record $86.2 billion that Elon Musk's SpaceX raised in its June IPO. SpaceX, which absorbed Musk's AI startup xAI before going public, made its listing a partial AI bet, too. Its shares skyrocketed initially before coming back down to earth, and now sit at about their $135 offering price. Anthropic's investors will likewise have to stomach huge losses for the foreseeable future. The company, according to US media reports, lost almost $42 billion in 2025, and will likely keep bleeding cash for years. To lure investors, according to the Wall Street Journal, Anthropic will promise that revenue opportunities are above $30 trillion.
Five things to know about Anthropic ahead of Wall Street debut
San Francisco (United States) (AFP) - Anthropic, the OpenAI rival that bet everything on computer coding, is expected to go public within weeks in a listing that could eclipse SpaceX's record Wall Street debut in June. Here are five things to know about the company: Built from OpenAI Anthropic was founded in 2021 by former OpenAI executives frustrated over how the potential of AI and concerns over safety were not understood or being taken seriously enough. The company -- whose name means, somewhat paradoxically, "relating to human beings" -- is led by CEO and co-founder Dario Amodei, a San Francisco native with a PhD in biophysics from Princeton University, not computer engineering like so many of his Big Tech peers. His sister Daniela is also a co-founder and the company's president. Anthropic has 5,000 employees, according to PitchBook. Upstart Until this year, Anthropic was in clear second place to OpenAI, which burst onto the scene with ChatGPT in November 2022, transforming the tech industry and triggering an AI arms race. But as OpenAI rolled out products from video creation to web browsers, Anthropic aimed far more narrowly, focusing on building the best platform for computer programmers. That strategy has paid off spectacularly -- coding is the rare AI skill that users are willing to pay handsomely for. Claude Code, its assistant for developers, has become one of the company's most popular products, helping push projected annual revenue to $65 billion. Only a small percentage of ChatGPT users, meanwhile, pay a subscription fee -- and OpenAI has put video AI and other side projects on the back burner. Trump vs. Anthropic The momentum comes despite severe political headwinds, with Anthropic at loggerheads with the Trump administration -- a state of affairs that could give investors pause. In March, the government broke off its contracts with Anthropic and designated the company a supply chain risk after it refused to grant the military unfettered access to its AI models. Anthropic called the Defense Department's move unconstitutional retaliation, and the two sides are now locked in a legal battle that could take years to resolve. The White House also bristles at Amodei's repeated warnings about the dangers of AI -- including the impact on jobs -- and his calls to regulate its deployment like airlines or banks. Amodei is also linked with effective altruism, a philosophy of targeted charitable giving scorned by conservatives in Silicon Valley and Washington. Big money needed Like OpenAI, Anthropic has massive needs for the computing power and infrastructure required to build so-called frontier models that stay ahead of competitors, amid fears that China could catch up. Both have raised money at unprecedented levels, with Anthropic valued at just under one trillion dollars when it raised $65 billion in May. With venture capitalists -- and even sovereign wealth funds -- no longer able to play in that league, higher sums can only be found on the public markets. This is a risky step that will probably determine whether a business model exists for Anthropic's vision of the AI revolution. OpenAI, after initially promising an IPO this year as well, is now signaling it will wait until 2027. Stomaching the losses According to Bloomberg, Anthropic intends to surpass the record $86.2 billion that Elon Musk's SpaceX raised in its June IPO. SpaceX, which absorbed Musk's AI startup xAI before going public, made its listing a partial AI bet, too. Its shares skyrocketed initially before coming back down to earth, and now sit at about their $135 offering price. Anthropic's investors will likewise have to stomach huge losses for the foreseeable future. The company, according to US media reports, lost almost $42 billion in 2025, and will likely keep bleeding cash for years. To lure investors, according to the Wall Street Journal, Anthropic will promise that revenue opportunities are above $30 trillion.
Anthropic Presents 30 Trillion Dollar TAM, Surpassing SpaceX
AI startup's IPO targets $100B raise, $2T valuation citing vast market potential The Wall Street Journal (WSJ), citing sources, reported on the 25th (local time) that Anthropic is expected to present a total addressable market (TAM) of over 30 trillion dollars (approximately 4 quadrillion 1,500 trillion Korean won) in its investment prospectus for its initial public offering (IPO). This exceeds the 28.5 trillion dollars proposed by Elon Musk's SpaceX in June. At the time, SpaceX's TAM included 26.5 trillion dollars allocated to its AI sector. SpaceX categorizes xAI, the AI model 'Grok,' and the social media platform X (formerly Twitter) under its AI business. TAM estimates the theoretical annual revenue a company could generate if it captured 100% of a market. As a metric for gauging future market size, some level of estimation is inevitable. This uncertainty is amplified in the AI industry, where growth speed and scope of application are difficult to predict. SpaceX's TAM announcement also drew skepticism on Wall Street due to its unusually large scale compared to existing cases. The combined revenue of 191 technology companies in the S&P 1500 Index last year amounted to only 2.4 trillion dollars. Aswath Damodaran, a finance professor at New York University known as a "valuation guru," told the WSJ that SpaceX's AI sector TAM figure was "beyond the bounds of plausibility." According to sources, Anthropic is currently reviewing all tasks that AI models could complete to calculate its TAM. Meanwhile, Anthropic is reportedly aiming to raise up to 100 billion dollars in its IPO, surpassing SpaceX's 86 billion dollars. Anthropic targets a corporate valuation of 2 trillion dollars, which would also exceed SpaceX's 1.77 trillion dollars based on its IPO price. Anthropic's IPO is expected to take place as early as September or early October.
Nvidia reportedly eyes another investment in Perplexity AI at a $30B valuation
Nvidia reportedly eyes another investment in Perplexity AI at a $30B valuation Nvidia Corp. is reportedly considering making another investment in the artificial intelligence search startup Perplexity AI Inc. A report by The Information says the chipmaker is holding talks with Perplexity over an investment that could push the startup's valuation to more than $30 billion. That would represent a jump of more than 50% from the $20 billion valuation Perplexity finalized about a year ago, when it last raised money. The size of Nvidia's potential investment was not disclosed, and there's no guarantee that any deal would be reached, The Information said, citing anonymous sources who are familiar with the discussions. Neither Nvidia nor Perplexity would comment on the reported discussions. Perplexity is an attractive target for investors for its business has continued to grow at a rapid rate. According to The Information, the startup has grown its annualized revenue run rate to an impressive $750 million, up from less than $250 million at the start of the year. If true, that would mean it has managed to triple its revenue run rate in just eight months. One of the main reasons for that impressive growth is Perplexity Computer, a cloud-based AI agent that was first released in April for Mac computers and later expanded to Windows devices. Perplexity Computer is designed to automate computer tasks for professional users. It acts as a general-purpose digital worker that can access authorized files and applications on a user's computer. Users can ask it to create or edit Word documents, update Excel spreadsheets, organize files, conduct online research and complete workflows involving multiple applications. The proposed investment would deepen an existing relationship between Nvidia and Perplexity. The chipmaker is already one of its main financial backers, alongside Amazon.com Inc. founder Jeff Bezos and SoftBank Group Corp. Nvidia has become an increasingly important partner for AI startups like Perplexity, and sees its bet on the startup as an investment in its future. As the world's top supplier of silicon for high-frequency AI inference, it has a vested interest in making sure that the search layer - which is a massive compute ecosystem - remains aligned with its chip ecosystem. What Nvidia doesn't want is for the likes of Perplexity and others to go sniffing around rival chipmakers such as Advanced Micro Devices Inc. and Cerebras Systems Inc., which both offer alternative chips for AI inference. In that way, Nvidia is investing in Perplexity as a kind of insurance policy to safeguard its future revenue stream against possible shifts in AI search architecture. Perplexity's strategic importance to Nvidia is amplified by its distribution efforts, such as its integration with Samsung Electronics Co. Ltd.'s Bixby assistant, which brings its search capabilities to around 800 million devices globally. The AI search firm is also believed to be fixed on a 2028 initial public offering, which means Nvidia has a clear timeline to realize a return on its investment. Nvidia's broader portfolio includes many of its major compute customers, including OpenAI Group PBC, Anthropic PBC, SpaceX Corp.'s xAI, Poolside Inc. and Safe Superintelligence Inc. It shows how the chipmaker has taken a systematic approach to ensuring its market dominance. By supplying the critical infrastructure and acting as a key investor at the application layer, Nvidia has effectively built a self-reinforcing cycle of demand for its chips. Nvidia is also trying to provide direct funding to customers that need to invest in its AI compute hardware. It recently struck a deal with six of Wall Street's biggest financial institutions to provide more than $500 billion in financing for AI infrastructure projects, including its own and those of its customers. Meanwhile, Perplexity has been racing to build out the infrastructure foundation it needs to support its own growth. Earlier this year, it struck a $750 million deal with Microsoft Corp. that will see it adopt that company's Azure cloud services to help run its AI workloads.
Anthropic Is Worth Much Less Than SpaceX
Anthropic is a pure play in what appears to be the faltering AI space. At least SpaceX (NASDAQ: SPCX | SPCX Price Prediction) has a rocket and an internet division. Recently, it became clear that at least some large corporations are willing to take slightly less AI firepower than Anthropic offers. And, by the way, Anthropic has tens of millions of dollars of obligations for data centers and Nvidia's (NASDAQ: NVDA) chips. (For some reason, Nvidia always seems to come out on top in all of these transactions.) A large group of investors believes an Anthropic IPO will top the record set by SpaceX. The main reasons are revenue and the fact that AI may be the most important invention in human history. Anthropic is outpacing its rivals' revenue run rates, which is one reason it is such a hot investment. The estimated run rate for this year is $65 billion; OpenAI's is as low as $40 billion based on the same calculation. The primary reason is corporate and institutional adoption. Anthropic's Claude has become the industry leader. Investors don't want to see individuals running Claude on their laptops. They want to see its AI functions at the world's largest companies because that is where the real money is. But the FT recently ran a headline that read, "Anthropic's best AI model struggles to attract users as cheaper tools thrive." This happens at the same time inexpensive Chinese models are rushing into the market. No one could have anticipated this Chinese surge even a year ago. Only days ago, newspapers reported that Anthropic has also largely dodged the concern that AI is just too expensive, even for large companies. However, some others have said the investment has not yielded a strong ROI and have cut back spending, at least temporarily. For "temporarily" to go away, AI ROI has to improve significantly. The SpaceX IPO gave the company a valuation of $1.77 trillion on the first trading day, and it raised $86 billion. That money is disappearing quickly and has gone to the SpaceX xAI division. So why the enthusiasm for Anthropic compared to SpaceX? SpaceX has the rocket business cornered. Its Starlink should become the de facto internet access for most of the world. However, its AI business is not attractive, even a little. Even with capex of $18.4 billion in the most recent quarter, it is not enough. Elon Musk, SpaceX CEO, said the capex sum must be much, much bigger. He needs more and more AI data center capacity. However, his models benchmark much behind those from Anthropic, OpenAI, and Google, at least. And then,, again, there are the Chinese. But if you look at the bets an investor takes, a shareholder in SpaceX is betting on three divisions. An investor in Anthropic is only looking at one. Anthropic is an AI pure play, the sector's consensus leader. SpaceX was a rocket and internet business with an AI business bolted on. That means that the IPO value of SpaceX won't be topped by Anthropic. Contact [email protected] for any questions or corrections.
Anthropic Expects Its IPO to Match or Beat SpaceX's $86 Billion Record - Memeburn
Users on X are questioning whether the product justifies the valuation, citing rate limits, invisible watermarks and pricing Anthropic is preparing what could become the largest initial public offering in history. The Claude developer expects its upcoming IPO to match or exceed the $86.2 billion that SpaceX raised when it went public in June, Bloomberg reported on August 20, citing people familiar with the matter. The company could file its IPO paperwork publicly as soon as the end of this month. If it clears that mark, 2026 would set the record for total US IPO volume, with companies that debuted on public markets already bringing in $160.6 billion as of August 19. But the mega-listing arrives alongside a quieter, more telling shift. Anthropic is also walking back a controversial data retention policy that had alarmed enterprise customers for months. And on social media, users are asking a blunter question: does the product actually justify a valuation this large? The Numbers Behind the Record Attempt Morgan Stanley, Goldman Sachs and JPMorgan Chase are managing Anthropic's offering, the same banks that ran SpaceX's blockbuster listing earlier this year. SpaceX initially targeted $75 billion but ultimately raised $86.2 billion once its overallotment option was exercised. Anthropic's financial trajectory is the core of the pitch. Its annualized revenue run rate hit $65 billion by late July, up from $9 billion at the end of 2025. That is more than a sevenfold increase in roughly seven months. Preliminary second-quarter revenue exceeded $11.5 billion, and investors told the Financial Times they expect annualized revenue to land between $100 billion and $120 billion by year-end. The company also reported its first positive adjusted operating income during Q2. Chief Financial Officer Krishna Rao has led recent investor briefings but has declined to discuss specific valuation targets, according to Bloomberg's sources. None of this came cheap. Anthropic posted a net loss of nearly $42 billion in 2025, driven by enormous compute costs. It has committed $50 billion to AI infrastructure across data centres in Texas and New York. CEO Dario Amodei has publicly acknowledged the stakes, warning that even a one-year miss on growth could threaten the company's survival. Some backers have floated a potential $2 trillion listing valuation by October, which would dwarf SpaceX's $1.77 trillion debut. But the IPO raise itself the amount of capital Anthropic actually collects from selling shares is the more immediate benchmark. Getting past $86.2 billion would give Anthropic the outright record. Anthropic Reverses Course on Data Retention The IPO news broke alongside a separate Bloomberg report: Anthropic plans to let enterprise customers store data on their own cloud infrastructure rather than Anthropic's servers. The shift matters because of what came before it. When Anthropic launched Claude Fable 5 in June, it introduced a mandatory 30-day data retention policy for its most powerful models. Every prompt, every output, stored for a month with no opt-out. The policy was designed to help the company detect misuse and monitor for cybersecurity threats, but enterprise customers pushed back hard. Microsoft reportedly restricted employee use of Anthropic's latest models while reviewing the policy. Salesforce and more than 100 other customers spent months working with Anthropic on alternatives. The revised approach keeps the 30-day retention requirement but gives businesses the option to host that data within their existing cloud setup. Anthropic says it expects to roll out the new safety system later this year. The timing is not a coincidence. Palantir CEO Alex Karp had publicly criticized AI companies for what he described as a data grab. OpenAI responded first, previewing its own zero-retention safety processing system for enterprise customers. Anthropic's reversal followed within 24 hours. For a company about to ask public investors for $86 billion, appearing to cave on data privacy under competitive pressure is a story the S-1 roadshow would rather not have running in the background. Users Aren't All Buying It While investors prepare for the biggest AI IPO ever, some of Anthropic's own users are greeting the news with scepticism. One widely shared post on X from user @jumperz summed up the frustration in a list: expensive flagship model, rate limits that cut users off mid-workflow, an invisible text watermark embedded in outputs, and a product they described as "out of touch with what users actually want." The replies leaned into it. One user said their first move would be to short the stock. Another compared the company unfavourably to SpaceXAI, pointing out that Elon Musk's combined entity launches reusable rockets and runs a global satellite internet service on top of building AI. "Why would you ever invest in Anthropic over SpaceX?" they asked. These are social media reactions, not institutional analysis. But they reflect a gap that Anthropic will need to address once it becomes a public company: the distance between what growth-stage investors see in the revenue numbers and what daily users experience in the product. Anthropic has dealt with product controversies before. Its export control clash with the White House in June forced it to disable its most capable models worldwide for weeks. Its multi-agent testing revealed AI coordination failures that made headlines. Each incident chipped at the narrative of a company that moves carefully and gets things right. An IPO prospectus will lay bare the financials. What it won't resolve is whether the product experience matches the price tag. What Happens Next Anthropic is running financial analyses ahead of a potential public filing by the end of August. If the company does file, the S-1 will contain the first public disclosure of detailed revenue, costs, and operating losses, figures that until now have come only through investor briefings and media leaks. The IPO is expected to list on the Nasdaq, with an October 2026 debut as the most likely timeline. OpenAI, which filed its own confidential S-1 in June, may delay its listing to 2027, potentially giving Anthropic a clear window to dominate public market attention. For context, SpaceX's own post-IPO run was anything but smooth. Shares surged 67% in four days, then crashed 49% over the following weeks before stabilising. AI IPO investors should expect similar volatility. The question is no longer whether Anthropic will go public. It is whether an AI company that lost $42 billion last year, changed its data policy under pressure, and is drawing public complaints about its product can convince both Wall Street and Main Street that the growth curve justifies the record. FAQs Will Anthropic's IPO be bigger than SpaceX's? Bloomberg reports that Anthropic is targeting an IPO raise at or above the $86.2 billion SpaceX collected in June 2026. SpaceX currently holds the record for the largest first-time share sale in history. Clearing that figure would give Anthropic the outright record. When is Anthropic expected to go public? Anthropic could file its IPO paperwork publicly as soon as late August 2026, with an October 2026 listing on the Nasdaq as the most likely target. Morgan Stanley, Goldman Sachs and JPMorgan Chase are managing the offering. What is Anthropic's current revenue? The company's run rate reached $65 billion by late July 2026, up from $9 billion at the end of 2025. Backers project that figure will land between $100 billion and $120 billion by year-end. Preliminary Q2 revenue exceeded $11.5 billion. Why did Anthropic change its data retention policy? Anthropic introduced a mandatory 30-day data retention policy for its most capable AI models in June 2026. Enterprise customers objected, and the company responded in August by offering an alternative: businesses can now retain the required data within their own cloud environments rather than handing it to Anthropic. Is Anthropic profitable? Not yet on an annual basis. The company's 2025 financials showed heavy losses driven by compute spending. However, Anthropic posted its first positive adjusted operating income in Q2 2026, and investors project annualized revenue could reach $100 billion or more by year-end, suggesting a path toward sustained profitability is forming.
