News & Updates
The latest news and updates from companies in the WLTH portfolio.
Is Anthropic's Mythos a genuine threat to your bank?
Australia and South Korea are watching Anthropic's Mythos AI over banking security risks. A powerful artificial intelligence model is drawing urgent attention from financial authorities across the globe, and the concerns being raised go well beyond routine tech oversight. Anthropic's frontier AI model Mythos is now under active scrutiny from regulators in Australia and South Korea, who are evaluating what the technology could mean for the stability of the world's banking systems, according to reporting from Reuters. Why Mythos has regulators on edge The concern is rooted in what experts describe as Mythos's exceptionally advanced ability to write and analyze code at a high level. That capability, they say, gives the model a potentially unprecedented ability to identify cybersecurity vulnerabilities within financial system infrastructure, raising the prospect that it could be used to exploit weaknesses in ways that no previous AI model has been positioned to do. It is that combination of coding sophistication and scale that has pushed Mythos into the crosshairs of regulators who typically move slowly on emerging technology. Three concerns driving regulatory action Experts and regulators have pointed to three compounding risks that make Mythos a particular focus of attention. First, its high-level coding capability sets it apart from earlier AI models, giving it the technical reach to probe complex financial systems. Second, the potential to identify cybersecurity vulnerabilities at scale means the threat is not theoretical but practically reachable by bad actors who might leverage the model. Third, the speed at which frontier AI is advancing leaves regulators concerned that existing frameworks may not be equipped to respond quickly enough if something goes wrong. Australia's financial regulators respond Two of Australia's most prominent watchdogs have confirmed they are actively monitoring the situation. The Australian Securities and Investments Commission has said it is tracking developments alongside peer regulators globally to assess any implications for the Australian market, while remaining in close contact with government agencies and the financial sector. The Australian Prudential Regulation Authority has similarly stated it is evaluating the broader implications of recent technological advances to safeguard the ongoing safety and resilience of the country's financial system. South Korea moves faster South Korea's response has been more immediate. The Financial Supervisory Service held a dedicated meeting with information security officials from major financial firms to assess risks linked directly to Mythos. The country's Financial Services Commission went further, convening an emergency session that brought together chief information security officers from banks and insurers to evaluate the situation. Those meetings signal a level of urgency that goes beyond standard monitoring. What comes next No formal regulatory action has been announced, and Anthropic has not issued a public response to the scrutiny. The fact that multiple regulators across separate countries are independently examining the same model at the same time, however, marks a meaningful moment in how governments are beginning to treat frontier AI as a direct concern for critical financial infrastructure. That conversation is almost certainly just getting started.
Cerebras Stock: Trying To Disrupt Nvidia's Inference Monopoly (Pending:CBRS)
Looking for a helping hand in the market? Members of Beyond the Wall Investing get exclusive ideas and guidance to navigate any climate. Learn More " New Details On Cerebras Systems Ahead Of Its IPO I wrote about Cerebras Systems (CBRS) in October 2024, when we first saw the initial IPO details, and back then, I called the offering's details quite Oakoff Investments is a personal portfolio manager and a quantitative research analyst with 5 years helping readers find a reasonable balance between growth and value by sharing proprietary Wall Street information. He leads the investing group Beyond the Wall Investing with features that include: a fundamentals-based portfolio, weekly analysis on insights from institutional investors, regular alerts for short-term trade ideas based on technical signals, ticker feedback by request from readers, and community chat. Learn more. Analyst's Disclosure: I/we have no stock, option or similar derivative position in any of the companies mentioned, and no plans to initiate any such positions within the next 72 hours. I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article. Seeking Alpha's Disclosure: Past performance is no guarantee of future results. No recommendation or advice is being given as to whether any investment is suitable for a particular investor. Any views or opinions expressed above may not reflect those of Seeking Alpha as a whole. Seeking Alpha is not a licensed securities dealer, broker or US investment adviser or investment bank. Our analysts are third party authors that include both professional investors and individual investors who may not be licensed or certified by any institute or regulatory body. CBRS postponed its IPO due to stagnating equity capital markets following 2022, which slowed new issuances across the sector. Cerebras uses wafer-scale processors that reduce latency and improve efficiency, offering an alternative architecture to traditional GPU clusters used by Nvidia. Its transition to an AI cloud provider could unlock recurring revenue streams and higher margins, significantly enhancing long-term profitability.
Anthropic's MCP vulnerability: When 'expected behavior' becomes a supply chain nightmare - TechTalks
An architectural vulnerability baked into the core of Anthropic's Model Context Protocol (MCP) exposes millions of AI applications to remote command execution. Security researchers at OX Security discovered a fundamental flaw in how the protocol handles local process execution, allowing attackers to hijack servers, exfiltrate private data, and infiltrate enterprise networks. Because the vulnerability exists at the protocol layer, the blast radius is massive. It affects over 150 million downloads, leaves more than 200,000 public servers potentially exposed, and has resulted in over 10 Common Vulnerabilities and Exposures (CVEs). The research team successfully executed commands on six live production platforms with paying customers and bypassed security checks on 9 out of 11 major MCP marketplaces. The vulnerability mostly remains in the wild as it is a "feature not a bug," and requires vigilance by developers. But the report highlights the kind of care you need to put into your AI applications as you adopt new technologies. To understand the scope of the threat, you have to understand the role of MCP. Released by Anthropic in November 2024, MCP acts as a universal plug adapter for AI agents. Large language models (LLMs) cannot inherently browse your local files or query a private SQL database. MCP bridges this gap. An MCP adapter translates the LLM's requests into actions the external service such as a web search engine or a database can understand. Because MCP sits at the junction between the AI model and external tools, it holds highly privileged access to source code, chat logs, and API keys. If the command successfully starts an MCP connection, the system proceeds. However, if the command is malicious and fails to establish a valid MCP connection, the system simply returns an error to the user. The critical failure is that the underlying operating system command still runs. There are no sanitization warnings or roadblocks. An attacker can pass a malicious command, receive a connection error, and walk away with full control of the server. The architecture makes it easy to distribute exploits. Developers frequently browse community marketplaces to download custom MCP configurations. OX Security uploaded a harmless proof-of-concept payload to 11 directories. Nine of them, including popular hubs like LobeHub and Cursor Directory, published the payload without any security review (GitHub was the only one that rejected the submission, and Cline did not respond). A single malicious listing could grant an attacker hidden access to thousands of developer machines. In practice, this flaw manifests differently depending on the application environment. LangFlow, an open-source framework owned by IBM, allows users to build AI workflows through a web interface. Researchers found over 915 publicly accessible LangFlow instances online. The platform exposes its MCP configuration panel without requiring authentication. Anyone on the internet can request a session token, send a crafted network request containing a malicious STDIO command, and take over the underlying server without ever logging in. The threat also extends to local development environments through prompt injection. AI-powered Integrated Development Environments (IDEs) like Windsurf and Cursor read local files and browse the web to assist developers. If a developer visits an attacker-controlled website, the site can feed a hidden, malicious instruction to the AI assistant. The AI then proposes an edit to the local mcp.json file to include a dangerous STDIO command. In the case of Windsurf, the system applied the edit and executed the command immediately, requiring zero clicks or approvals from the developer. When OX Security reported these findings, the response from protocol maintainers and IDE vendors was uniform. Anthropic, LangChain, and Microsoft all stated that the behavior is expected and operates by design. The core of their argument is flexibility. To maintain MCP as an unopinionated, universal standard, the protocol avoids putting strict limitations on what developers can execute. LangChain maintains that application authors are responsible for validating and sanitizing inputs. IDE vendors argue that users must actively trust their workspace environments. This approach creates a structural failure. It shifts the burden of complex security sanitization onto tens of thousands of downstream developers building chat apps and internal tools. Many of these developers are not security engineers, guaranteeing that vulnerabilities will emerge at scale. Moshe Siman Tov Bustan, security research team lead at OX Security, argues that safety does not require sacrificing functionality. "The protocol can be redesigned in a way that doesn't lose its utility," Bustan told TechTalks. "By letting developers choose the available functionality, it could block arbitrary commands by default, but let developers enable them by passing a flag that clearly states the function enables arbitrary command execution and should not be exposed to user input." Implement manifest-only execution: The pattern of passing raw, user-supplied strings directly into a shell execution environment must stop. Teams should transition to a manifest model, where applications only execute pre-defined server aliases locked within a static configuration file. Enforce strict sandboxing: AI agents should operate in low-privilege environments. "If an MCP needs to communicate with a remote server, it should be allowed to contact only that specific server. The same applies to filesystem access and any other behavior," advises Bustan. Require explicit opt-ins: Developers should introduce a mandatory flag whenever dynamic STDIO arguments are active. This makes dangerous modes easily searchable for security linters and audit teams. Adopt least-privilege secret management: API keys and OAuth tokens provided to MCP servers must be strictly scoped to their intended function. The presence of broad, full-permission keys should trigger automated warnings during the deployment process. Establish marketplace verification: Registries need to mature beyond open uploads. Marketplaces should require developers to submit a standardized security manifest detailing exactly which network resources and file systems an MCP server will access before allowing users to download it. The vulnerabilities within MCP highlight a broader shift in how software is built today. The widespread use of AI coding assistants and vibe-coding tools allows individuals with limited security maturity to generate and deploy functional code at an unprecedented speed. OX Security refers to this dynamic as the "Army of Juniors" effect. AI systems are highly effective at writing integration logic, but they rarely question the architectural safety of the underlying protocols they use. "We consider this cascading vulnerability to align with this AI coding anti-pattern, because AI services are often written by AI-native developers," Bustan said. "If the AI doesn't question the underlying implementation, the developer behind the AI won't either." This dynamic makes foundational security critical. Protocol maintainers must embed "secure by design" principles directly into SDKs and package dependencies, rather than shifting the burden of sanitizing inputs downstream. As autonomous systems gain more access to sensitive data and critical infrastructure, relying on traditional tools or even other AI systems to catch anomalies is no longer enough. "While code is rapidly evolving and being written at scale, security expertise and manual reviews are now needed more than ever," Bustan said. "Our analysis shows that we cannot blindly trust automated AI reviews to check code security. Especially in areas where the code handles sensitive information, human manual reviews can identify architectural flaws that are often introduced by AI agents."
Vandenberg rocket launch schedule. SpaceX missions due for liftoff
Here's your weekly look at the upcoming launch schedule at the Vandenberg Space Force Base in Santa Barbara County. Two rocket launches are scheduled for the week ahead from the Vandenberg Space Force Base in California. And both of the missions will see SpaceX launch its Falcon 9 rocket to deploy two more batches of the commercial spaceflight company's Starlink broadband internet satellites to Earth orbit. Californians - and potentially Arizonans - making their plans to watch any launches should keep in mind that lots can change between now and when each rocket is expected to lift off, as delays in spaceflight are common. Be sure to visit VC Star for the latest mission updates. In the meantime, here's your weekly look at the upcoming launch schedule at the Vandenberg Space Force Base in Santa Barbara County. Wednesday, April 22: Starlink 17-14 * Agency: SpaceX * Mission: Deploy 25 Starlink satellites into low-Earth orbit, according to SpaceX * Rocket: Falcon 9, a 230-foot, two-stage reusable rocket capable of carrying more than 50,000 pounds of cargo to low-Earth orbit * Launch window: 7 to 11 p.m. PT Wednesday, April 22, 2026 * Rocket launch location: Space Launch Complex 4E from Vandenberg Space Force Base in California * Booster landing: SpaceX drone ship, nicknamed "Of Course I Still Love You," in the Pacific Ocean Saturday, April 25: Starlink 17-16 * Agency: SpaceX * Mission: Deploy 25 Starlink satellites into low-Earth orbit, according to SpaceX * Rocket: Falcon 9, a 230-foot, two-stage reusable rocket capable of carrying more than 50,000 pounds of cargo to low-Earth orbit * Launch window: 7 to 11 a.m. PT Saturday, April 25, 2026 * Rocket launch location: Space Launch Complex 4E from Vandenberg Space Force Base in California * Booster landing: SpaceX drone ship, nicknamed "Of Course I Still Love You," in the Pacific Ocean Eric Lagatta is the Space Connect reporter for the USA TODAY Network. Reach him at [email protected]
Vandenberg rocket launch schedule. SpaceX missions due for liftoff
Here's your weekly look at the upcoming launch schedule at the Vandenberg Space Force Base in Santa Barbara County. Two rocket launches are scheduled for the week ahead from the Vandenberg Space Force Base in California. And both of the missions will see SpaceX launch its Falcon 9 rocket to deploy two more batches of the commercial spaceflight company's Starlink broadband internet satellites to Earth orbit. Californians - and potentially Arizonans - making their plans to watch any launches should keep in mind that lots can change between now and when each rocket is expected to lift off, as delays in spaceflight are common. Be sure to visit VC Star for the latest mission updates. In the meantime, here's your weekly look at the upcoming launch schedule at the Vandenberg Space Force Base in Santa Barbara County. Wednesday, April 22: Starlink 17-14 * Agency: SpaceX * Mission: Deploy 25 Starlink satellites into low-Earth orbit, according to SpaceX * Rocket: Falcon 9, a 230-foot, two-stage reusable rocket capable of carrying more than 50,000 pounds of cargo to low-Earth orbit * Launch window: 7 to 11 p.m. PT Wednesday, April 22, 2026 * Rocket launch location: Space Launch Complex 4E from Vandenberg Space Force Base in California * Booster landing: SpaceX drone ship, nicknamed "Of Course I Still Love You," in the Pacific Ocean Saturday, April 25: Starlink 17-16 * Agency: SpaceX * Mission: Deploy 25 Starlink satellites into low-Earth orbit, according to SpaceX * Rocket: Falcon 9, a 230-foot, two-stage reusable rocket capable of carrying more than 50,000 pounds of cargo to low-Earth orbit * Launch window: 7 to 11 a.m. PT Saturday, April 25, 2026 * Rocket launch location: Space Launch Complex 4E from Vandenberg Space Force Base in California * Booster landing: SpaceX drone ship, nicknamed "Of Course I Still Love You," in the Pacific Ocean Eric Lagatta is the Space Connect reporter for the USA TODAY Network. Reach him at [email protected]
Vandenberg rocket launch schedule. SpaceX missions due for liftoff
Here's your weekly look at the upcoming launch schedule at the Vandenberg Space Force Base in Santa Barbara County. Two rocket launches are scheduled for the week ahead from the Vandenberg Space Force Base in California. And both of the missions will see SpaceX launch its Falcon 9 rocket to deploy two more batches of the commercial spaceflight company's Starlink broadband internet satellites to Earth orbit. Californians - and potentially Arizonans - making their plans to watch any launches should keep in mind that lots can change between now and when each rocket is expected to lift off, as delays in spaceflight are common. Be sure to visit VC Star for the latest mission updates. In the meantime, here's your weekly look at the upcoming launch schedule at the Vandenberg Space Force Base in Santa Barbara County. Wednesday, April 22: Starlink 17-14 * Agency: SpaceX * Mission: Deploy 25 Starlink satellites into low-Earth orbit, according to SpaceX * Rocket: Falcon 9, a 230-foot, two-stage reusable rocket capable of carrying more than 50,000 pounds of cargo to low-Earth orbit * Launch window: 7 to 11 p.m. PT Wednesday, April 22, 2026 * Rocket launch location: Space Launch Complex 4E from Vandenberg Space Force Base in California * Booster landing: SpaceX drone ship, nicknamed "Of Course I Still Love You," in the Pacific Ocean Saturday, April 25: Starlink 17-16 * Agency: SpaceX * Mission: Deploy 25 Starlink satellites into low-Earth orbit, according to SpaceX * Rocket: Falcon 9, a 230-foot, two-stage reusable rocket capable of carrying more than 50,000 pounds of cargo to low-Earth orbit * Launch window: 7 to 11 a.m. PT Saturday, April 25, 2026 * Rocket launch location: Space Launch Complex 4E from Vandenberg Space Force Base in California * Booster landing: SpaceX drone ship, nicknamed "Of Course I Still Love You," in the Pacific Ocean Eric Lagatta is the Space Connect reporter for the USA TODAY Network. Reach him at [email protected]
Vandenberg rocket launch schedule. SpaceX missions due for liftoff
Here's your weekly look at the upcoming launch schedule at the Vandenberg Space Force Base in Santa Barbara County. Two rocket launches are scheduled for the week ahead from the Vandenberg Space Force Base in California. And both of the missions will see SpaceX launch its Falcon 9 rocket to deploy two more batches of the commercial spaceflight company's Starlink broadband internet satellites to Earth orbit. Californians - and potentially Arizonans - making their plans to watch any launches should keep in mind that lots can change between now and when each rocket is expected to lift off, as delays in spaceflight are common. Be sure to visit VC Star for the latest mission updates. In the meantime, here's your weekly look at the upcoming launch schedule at the Vandenberg Space Force Base in Santa Barbara County. Wednesday, April 22: Starlink 17-14 * Agency: SpaceX * Mission: Deploy 25 Starlink satellites into low-Earth orbit, according to SpaceX * Rocket: Falcon 9, a 230-foot, two-stage reusable rocket capable of carrying more than 50,000 pounds of cargo to low-Earth orbit * Launch window: 7 to 11 p.m. PT Wednesday, April 22, 2026 * Rocket launch location: Space Launch Complex 4E from Vandenberg Space Force Base in California * Booster landing: SpaceX drone ship, nicknamed "Of Course I Still Love You," in the Pacific Ocean Saturday, April 25: Starlink 17-16 * Agency: SpaceX * Mission: Deploy 25 Starlink satellites into low-Earth orbit, according to SpaceX * Rocket: Falcon 9, a 230-foot, two-stage reusable rocket capable of carrying more than 50,000 pounds of cargo to low-Earth orbit * Launch window: 7 to 11 a.m. PT Saturday, April 25, 2026 * Rocket launch location: Space Launch Complex 4E from Vandenberg Space Force Base in California * Booster landing: SpaceX drone ship, nicknamed "Of Course I Still Love You," in the Pacific Ocean Eric Lagatta is the Space Connect reporter for the USA TODAY Network. Reach him at [email protected]
Adobe partners with Microsoft, Anthropic, OpenAI, Nvidia: WSJ
Adobe is partnering with Microsoft, Anthropic, OpenAI, and NVIDIA, per the Wall Street Journal. The "largest company by June 30" market on Polymarket prices NVIDIA at YES. The odds for NVIDIA's market cap dominance by June 30 sit at YES, up from 84% yesterday. The market resolves in 73 days, and order book depth shows $9,363 needed to shift the price by 5 points. The second-largest by April 30 market sits at just YES. The gap between the two markets shows traders expect NVIDIA to hold the top spot by June but see almost no chance it drops to second by April. The June market's liquidity of $6,703/day means single large transactions can move the price meaningfully. The April second-largest market trades only $11 in actual USDC daily, making it essentially illiquid. Buying YES at pays $1 if NVIDIA is the largest company by June 30, a return. The April market offers no practical trading opportunity at current volume. Watch NVIDIA's upcoming earnings and any announcements from Microsoft or OpenAI about joint AI projects. These could move odds on the June market.
UFO Takes Flight: Agility and Assets Converge in the Final Frontier
While some investment themes have struggled to maintain their trajectory, the space economy has proven it has the fundamental fuel to reach escape velocity. This momentum is being met with a strategic modernization of the underlying infrastructure for UFO. VettaFi announced today that effective May 15, the S-Network Space Index -- the benchmark for the Procure Space ETF (UFO) -- will be rebranded as the VettaFi Space Index. This update also includes a "fast-track" methodology designed to bypass traditional seasoning periods for newly public industry leaders. By allowing for the immediate, day-one inclusion of mega-cap IPOs, the index will be structured to capture the full valuation cycle of transformative companies. Specifically, this update paves the way for a potential SpaceX addition in mid-2026, which current market projections estimate could debut with a valuation exceeding $1.5 trillion. "The space industry is no longer just about 'what if' -- it is about 'what is happening now,'" explained Jane Edmondson, Head of Index Product Strategy at VettaFi. "By fast-tracking the inclusion of dominant forces like SpaceX, we are ensuring that the VettaFi Space Index accurately reflects the leaders of the orbital economy, providing a more precise benchmark for the next generation of space innovation." The methodology shift arrives as UFO has emerged as a standout among thematic peers. The seven year old fund has more than tripled in size this year, recently crossing the $500 million asset milestone. The ETF's performance has remained robust; while broader market indices have seen a choppy start to the year, UFO gained 35% year-to-date through April 15. UFO has strong record, with 14% annualized total return over the last five years. In this time, it has seen the emergence of new thematic peers. The introduction of the "fast-track" rule is a significant shift for the index's concentration profile. Under the current modified market-cap weighting, the portfolio is diversified across approximately 50 constituents. However, the inclusion of a mega-cap leader like SpaceX would fundamentally redistribute these weights. SpaceX's projected scale is driven by its Starlink satellite broadband revenue. It would likely enter the VettaFi index as a dominant top-tier holding. This would transition UFO from a basket of predominantly mid-cap innovators into a more barbell portfolio. UFO would likely be anchored by a trillion-dollar industry leader. Meanwhile, it maintains exposure to high-growth specialists in Earth observation and deep-space logistics. According to Edmondson, by refining filters to ensure companies derive significant revenue or strategic value from space-based activities, VettaFi is doubling down on thematic purity. UFO's crossing the half-billion-dollar mark was a key milestone. However, the updated methodology ensures the fund stays as dynamic as the industry it tracks. The ETF's index approach provides benefits of transparency and agility required to lead the next phase of the space race. UFO is one of many thematic ETFs to track a VettaFi index. The ALPS Medical Breakthroughs ETF (SBIO), the Range Nuclear Renaissance ETF (NUKZ), and the ROBO Global Robotics & Automation ETF (ROBO) are other examples of how VettaFi's rules-based index approach is helping investors navigate complex, high-growth themes. For more news, information, and analysis, visit VettaFi | ETF Trends
London Workers Brace for Travel Chaos in Latest Tube Strike
London commuters are facing significant travel disruption again this week with drivers on the city's Underground train network poised to walk out in a dispute over working hours. Staff on the Tube will push ahead with industrial action after talks failed on an agreement over Transport for London's plan to allow drivers to condense their five-day week into four days by working longer hours. Walkouts will take place over two 24-hour periods this week, with the first starting at 12 p.m. on Tuesday. More are planned for next month and also for June. Tube Driver Strike Dates:April 21, 12 p.m. (Tuesday) to April 22, 11:59 a.m. (Wednesday)April 23, 12 p.m. (Thursday) to April 24, 11:59 a.m. (Friday)May 19, 12 p.m. (Tuesday) to May 20, 11:59 a.m. (Wednesday)May 21, 12 p.m. (Thursday) to May 22, 11:59 a.m. (Friday)June 16, 12 p.m (Tuesday) to June 17, 11:59 a.m. (Wednesday)June 18, 12 p.m. (Thursday) to June 19, 11:59 a.m. (Friday)Source: RMT Two strikes planned last month were called off by the RMT labor union following progress in talks. But the chances of a truce appear limited this time round after the union criticized TfL at the weekend. "We have approached negotiations with TfL in good faith throughout this entire process," said RMT General Secretary Eddie Dempsey. "But despite our best efforts, TfL seem unwilling to make any concessions in a bid to avert strike action." The union has said its members are concerned about shift lengths and the potential impact on fatigue and safety, if drivers were to work longer daily shifts. These are the first strikes since September. The changes will help improve reliability and are voluntary for workers, TfL has said. Many London workers will work from home to avoid delays traveling to and from the office. That won't be much consolation for the central London bars and restaurants that are set to lose business. Pubs and bars may see a drop in sales as high as 40% during the walkouts, the UKHospitality trade body has predictedBloomberg Terminal, a fresh setback for a sector already grappling with higher wages and taxes.
Australia's banking regulator on Anthropic's Mythos: We expect financial services companies to ...
Australia's financial regulators have joined global counterparts in sounding the alarm over Anthropic's frontier AI model Mythos, which experts say could very well powerful enough to destabilise banking systems by exposing cybersecurity vulnerabilities at scale. According to a report by Reuters, the Australian Securities and Investments Commission (ASIC) said that it is closely monitoring Mythos along with other regulators. A spokesperson noted: "ASIC engages closely with other regulators, government agencies and the financial sector to understand and respond to changing technologies." The agency emphasised that that financial services licenses are expected to be proactive in safeguarding customers and clients against risks posed by advanced AI systems.The report by Reuters further adds that the Australian Prudential Regulation Authority (APRA) echoed ASIC's concerns, stating it will continue to assess the implications of technological advancements like Mythos to ensure the ongoing safety and resilience of the financial system.The warning from Australia is followed by similar moves in South Korea, where the Financial Supervisory Service (FSS) convened meetings with banks and insurers to review Mythos-related risks. Yonhap news agency reported that the Financial Services Commission (FSC) also held an emergency session with chief information security officers last week.Recently, Executives from Canada's largest banks and top regulators gathered to discuss the cybersecurity risks posed by Anthropic's new Claude Mythos AI model, amid the growing concerns that the technology could be weaponised to exploit software vulnerabilities. According to a report by The Globe and Mail, the meeting was held by the Canadian Financial Sector Resiliency Group (CFRG), chaired by Bank of Canada COO Alexis Corbett, and included representatives from the Department of Finance, the Office of the Superintendent of Financial Institutions (OSFI), and executives from Canada's six biggest banks plus Desjardins Group.As per the report, Bank of Canada spokesperson Paul Badertscher emphasised that the meeting was not an emergency one but rather a 'situational awareness' session. "It can still hold meetings at the request of its members. 'Hey guys, we need to pay attention, there is something going on. Let's get together and talk about this.' That's what this was," he said.
$292 Million KelpDAO Hack, Vercel Breach Highlight Rising AI-Driven Crypto Security Risks - TokenPost
A $292 million KelpDAO exploit and Vercel breach via an AI tool triggered major DeFi outflows and highlight escalating AI-driven security risks across crypto infrastructure. A pair of high-profile security incidents over April 18-19 has rattled both decentralized finance and mainstream cloud infrastructure, underscoring a growing reality: as AI accelerates software creation, it is also accelerating the pace and sophistication of attacks -- and the blast radius increasingly extends beyond smart contracts to the tools and services surrounding them. On April 18 UTC, KelpDAO's LayerZero-based cross-chain bridge was exploited, with an attacker siphoning roughly 116,500 rsETH -- valued at about $292 million -- within 46 minutes, according to figures cited in the Korean report. The stolen assets were quickly posted as collateral on Aave (AAVE) to borrow native Ether, a maneuver that amplified market stress by turning bridged exposure into immediate on-chain liquidity. Aave moved into an emergency freeze, triggering a broader DeFi shock. Total value locked (TVL) on Aave reportedly fell by about $6.6 billion, while AAVE dropped as much as 16% amid a rush to de-risk. The pullback spread quickly across protocols as users rushed to withdraw funds, with total outflows across DeFi estimated at around $10 billion -- behavior reminiscent of a digital 'bank run' driven by confidence rather than solvency. The following day, April 19 UTC, attention shifted from smart contracts to the web layer. Cloud platform Vercel -- widely used for hosting front ends and dashboards -- was breached via a third-party AI tool, Context.ai, used by an employee. The attacker allegedly used that foothold to seize access to a Google account and then move deeper into internal infrastructure. A post on a hacking forum claimed internal databases, employee accounts, and GitHub and NPM tokens were being offered for $2 million, the report said. For crypto teams, the Vercel incident carried an especially sharp lesson: the blockchain can remain intact while the user experience layer fails catastrophically. Smart contracts may continue to execute as designed, but compromised front ends can redirect transactions, steal credentials, or manipulate what users sign. In effect, the 'lock' may still work -- while the front door is left open. The incidents also fit into a wider pattern. The Korean commentary pointed to a string of April exploits -- ranging from Drift Protocol to CoW Swap, Zerion, Rhea Finance, and Silo Finance -- and said 2026 year-to-date DeFi losses had already surpassed $450 million across 45 protocols by mid-April. The takeaway is less about any single bug than about a structural change: attacks are becoming more frequent, more targeted, and faster to execute, shrinking the window for detection and response. Industry warnings are growing louder. Ledger's CTO was quoted as saying 2026 could become the worst year on record for hacking losses. While such statements are inherently speculative, the sequence of events illustrates why security teams are increasingly focused on systemic risk -- bridges, shared libraries, cloud services, identity providers, and AI-enabled workflows -- rather than isolated protocol code. At the center of the shift is AI itself. Anthropic CEO Dario Amodei has argued that AI could be writing the majority of code within months, and the broader market is already moving in that direction through copilots, agents, and automated deployment tools. The same capabilities that lower barriers to launching services -- rapid code generation, dependency selection, and automated configuration -- also lower barriers to offense, enabling faster vulnerability discovery, exploit construction, and social engineering at scale. The report cited internal testing of a recent Anthropic model that allegedly identified thousands of previously unknown vulnerabilities, including a flaw in a 27-year-old operating system component, and achieved a 72.4% success rate generating effective exploit code -- far above earlier models. Even if such figures vary by benchmark and environment, the direction is clear: AI is compressing the attacker's workflow from weeks to hours, while patch cycles, audits, and infrastructure modernization remain comparatively slow. Vercel's breach, in particular, highlights a 'new grammar' of attacks -- one that targets the connective tissue of modern development rather than a single codebase. The weak link was not necessarily Vercel's core product but an ecosystem of AI productivity tools, third-party OAuth relationships, and everyday employee accounts. In DeFi, the KelpDAO exploit similarly points to risk emerging at the seams -- how components are connected, how trust assumptions stack, and how quickly attackers can pivot from one layer to the next. The implications extend beyond any one region. Korean exchanges, financial institutions, and blockchain projects rely on the same global cloud providers, the same Linux-based infrastructure, and many of the same AI-enabled tools being adopted worldwide. Vulnerabilities do not respect borders, and dependence on shared platforms can turn a single compromised account or token into a multi-tenant event affecting many downstream services. For markets, the recent events reinforce a familiar but evolving conclusion: security risk is no longer confined to smart contract code. The next phase of crypto security will be increasingly shaped by 'identity,' 'front-end integrity,' and 'supply-chain trust' -- and by how quickly defenders can adapt processes to match the speed of AI-assisted attackers. The April incidents did not just expose bugs; they exposed a widening mismatch between modern composable systems and legacy assumptions about where security boundaries begin and end.
Ministers urge cyber crackdown as Anthropic AI fears grow
Ministers have ramped up pressure on UK businesses to strengthen their cyber defences amid concerns that the next generation of AI could supercharge hacking threats. In a new push led by cyber minister Baroness Lloyd of Effra, nearly 200 business leaders were told to sign a new "cyber resilience pledge", as fears mount that powerful AI tools could expose them at unprecedented speed. Her intervention follows the release of Anthropic's closely watched Mythos model, which triggered alarm across governments, banks and regulators alike. Early testing suggested it can identify and exploit software weaknesses far faster than a human could. Bank of England governor Andrew Bailey warned last week that the technology could "crack the whole cyber risk world open", whilst Baroness Lloyd pointed out: "AI is giving attackers capabilities that would have seemed extraordinary just a year ago and no organisation can afford to be complacent". Boardrooms on alert as risk escalates The government-backed pledge will require firms to treat cybersecurity as a board-level responsibility, adopt baseline protections like Cyber Essentials, and sign up to the National Cyber Security Centre's early warning systems. Ministers worry that many firms are unprepared, with only one per cent of UK businesses currently holding Cyber Essentials certification, despite repeated warnings on cyber risk. Meanwhile, testing by the UK's Security Insititute found Anthropic's Mythos model represents a significant leap forward, capable of carrying out multi-step cyber attacks and autonomously exploiting weak systems in controlled environments. According to the institute, these capabilities could soon be widely accessible - raising the risk of more frequent, faster and more sophisticated attacks on businesses. Security minister Dan Jarvis is expected to warn this week that cybercrime should be treated with the same seriousness as physical attacks. "If this damage had been caused by an old-school physical attack, it would have been the equivalent of hundreds of masked criminals... smashing up computers and driving cars right off the forecourt", he is set to say. The risks seem to already be materialising, with more than half of UK firms having been hit by state-backed cyber attacks last year, while nearly half say they have already experienced AI-powered attacks. The National Cyber Security Centre has also warned of a growing "severe cyber threat", where attacks could lead to prolonged outages, and major financial losses. Despite this, many companies lack the expertise or budget to deploy advanced defences, leaving a widening gap between the speed of attackers and the resilience of businesses. Ministers are now urging firms to act before the threat escalates further, warning that preparation cannot be improvised once a major incident hits. "Cyber-resilience isn't just a technical issue; it's a board responsibility," Lloyd said. "We're asking every boardroom in Britain to prove they treat it as one."
Lib Dem deputy leader says Reform would bring 'chaos' to county council
The deputy leader of the Lib Dems has warned voters of the potential "chaos" Reform UK could bring if the party claims County Hall at May's elections. Nigel Farage's party are tipped by some observers to take over Norfolk County Council next month, with pollsters predicting they could win as many as 66 of the 84 seats up for grabs. But Daisy Cooper, deputy leader of the Liberal Democrats, issued a message to people considering voting for Reform at a party rally in North Walsham on Friday. Sir Norman Lamb, North Norfolk MP Steffan Aquarone and Daisy Cooper at a Liberal Democrats rally on Friday (Image: Newsquest/LDRS) The MP for St Albans, who was born in Bury St Edmunds and went to school in Halesworth and Bungay before being privately educated at Framlingham College, said: "I think what I would say is Reform is clearly riding high in the polls, and I think the reason for that is people are sick to death of the Conservatives and feel the Conservatives failed not only in Westminster but in County Hall as well. "But what I would say to people is if they look at the Reform-led councils around the country, they are consumed with chaos and broken promises and scandal in many cases." According to the latest PollCheck forecasts, the Lib Dems - the main opposition to the Conservatives at County Council - are at risk of losing several seats in its North Norfolk heartland. Daisy Cooper, deputy leader of the Liberal Democrats (Image: Houses of Parliament) Nigel Farage's Reform UK are predicted to take control of County Hall (Image: Denise Bradley) They could hold just three wards after May 7 - Holt, Eaton and Costessey. The Norfolk Lib Dems have pledged to offer cheaper bus travel for younger people, use AI to fix potholes and create community hubs in towns for people to access key services should they be elected to County Hall. Ms Cooper said she was "encouraged" by the party manifesto, adding the Lib Dems can "restore trust" in politics. "There's a huge opportunity, but also responsibility, to demonstrate politics can deliver for people and that politics can change lives," the deputy leader continued. "If they look at the record of Liberal Democrats here in Norfolk, they can see well-run services, well-run district councils and voting for Liberal Democrats could help restore trust, protect local services and to make sure we look after taxpayers' money as well."
Column-Amid Middle East chaos, China changes tactics on Taiwan and Japan
April 17 (Reuters) - As U.S. Vice President JD Vance returned to Washington last weekend after unsuccessful peace talks in Pakistan over the crisis in the Gulf, China's Foreign Ministry was preparing for a bumper slate of visitors including Russian Foreign Minister Sergei Lavrov and the leaders of Spain, Vietnam and the United Arab Emirates. With the U.S. this week declaring its own military "blockade" of the Strait of Hormuz, and Iran continuing to demand cryptocurrency payments to guarantee vessels' safe passage, Beijing is stepping up diplomatic efforts to present itself as a wider global power broker. That includes presenting itself as a global voice of sanity and stability against a "regression to the law of the jungle", as Xi put it on Tuesday as he met Spanish Prime Minister Pedro Sanchez. But it also includes tough military and diplomatic messaging in Beijing's immediate environment, where it has further escalated growing tensions with Japan, moving vessels to prevent the Philippines from reaching a long-disputed shoal, and an assertive and very public outreach to Taiwan's opposition Kuomintang party ahead of 2028 elections. In the process, Beijing has subtly but swiftly redefined its primary Taiwan narrative and, perhaps even more importantly, its timescale for potential future action. Whereas U.S. officials have warned for years that Beijing has been building up its military with the explicit goal of being ready to invade by 2027, China's new narrative has refocused attention on Taiwan's election the next year, pushing harder than ever the concept that a Kuomintang victory would avert conflict and offer much closer relations. If the incumbent Democratic Progressive Party wins again, Beijing is now setting the tone for further confrontation - on one hand, pushing any future U.S. administrations to abandon the island to its fate; on the other, tearing up relations with Japan over its comments on the future of Taiwan. The reasons are not difficult to see. Official Chinese statements have blamed new Japanese prime minister Sanae Takaichi, particularly for comments in November that a Chinese takeover of Taiwan might constitute a "survival-threatening" situation for Japan. Chinese officials and news outlets are also expressing concern that Tokyo might look to acquire nuclear weapons - something that might make it almost impossible for China to risk an attack against Taiwan or indeed Japan itself. Takaichi's government says it remains committed to Japan's non-nuclear status. But it is keen to strengthen international relationships wherever possible, inviting more than 30 ambassadors from NATO and other nations to Tokyo from Wednesday to step up cooperation in defence and a host of other areas. On almost every other front, however, Beijing is making diplomatic inroads. In a heavily promoted meeting in Beijing last week, new Kuomintang leader Cheng Li-wun pledged to invite Xi to Taiwan if she wins the 2028 election. Meanwhile, the DPP government is working on its own defences, nervously eyeing a repeatedly postponed Trump-Xi summit for signs that the Chinese leader might press the U.S. to dial down its support for Taiwan, perhaps in return for help in calming the Gulf. BEIJING SEEKS 'CAREFUL LISTENERS' Exactly what Beijing is up to when it comes to Iran remains vague; Chinese officials and media outlets have so far angrily denied CNN reports that Beijing has given Tehran weapons in the current crisis. Other analysts note that, as when supporting Russia in Ukraine, China prefers to deliver "dual use" civilian components that can be used in arms manufacture rather than delivering whole weapons or systems. U.S. President Donald Trump has pledged to impose new tariffs on Beijing if arms shipments are proven, a move that would further damage already dwindling global growth. The prospect of letting the rest of the world blame the U.S. for the current global mayhem, meanwhile, is something China is grabbing with both hands. U.S. behaviour, not only in the Middle East but also over Greenland, has made much of Europe reappraise its relationship with Washington. European Union foreign policy supremo Kaja Kallas described U.S. actions in the Middle East as second only to Russia's invasion of Ukraine in its damage to the international system. Whether that prompts European and other nations to embrace Beijing, however, is an open question. Britain criticised Russia and China together for vetoing a U.N. Security Council resolution from Bahrain condemning Iran's obstruction of ships in Hormuz and demanding the strait be reopened. Trump's decision to impose his own blockade this week, however - with its implication that U.S. warships will intervene to stop Chinese and other vessels carrying Iranian and other energy - seems to have played into Beijing's hands, just as his import tariffs did, by allowing China to grab the agenda as a supporter of free trade. Some Chinese pundits now enthusiastically predict more of this to come: the Trump administration grabbing the headlines while, behind the scenes, a whole new architecture of international relations takes shape in which nations work around the U.S. to try to secure their vital interests. It is an environment in which many incidents and confrontations of considerable importance go almost unreported, from Pakistan's military clashes with Afghanistan to suspected Iranian drone strikes on energy infrastructure in Azerbaijan, a nation that has become increasingly critical to European energy supplies in recent years. "China is increasingly seen by many as a major country that has consistently emphasised peace, development, sovereignty and dialogue," Chinese international relations academic Mabel Miao Lu wrote in the Communist Party-run Global Times this week, after returning from the Baku Global Forum in Azerbaijan. "That does not mean all concerns about China have disappeared. It does mean that more people are listening carefully when China speaks." LESSONS TO LEARN FOR CHINA AND OTHERS The lesson Beijing would most like the rest of the world to learn might be that when or if China's military moves to take control of Taiwan - either because it has been invited in by the Kuomintang or because Xi's patience has finally run out - the rest of the world should not get involved, whatever Japan or the U.S. might do. At the same time, many countries will have concluded that action such as that taken by the U.S. in Iran, or Russia in Ukraine, triggers massive and immediate global dislocation - and, although the energy shock of conflict in the Gulf has been profound, the supply shocks of a major war in the Pacific could be far worse. While the U.S. may not have gained what it wanted from Operation Epic Fury, as with strikes on Tehran's nuclear programme last June, the Pentagon has delivered a lesson to other potential foes that the U.S. military retains a level of expertise in complex operations that Beijing has never trialled in anger. Xi's confidence in his own military may only be mixed, at best, judging by the number of senior military commanders removed in recent months and years - although China's sheer manufacturing capacity may still keep Pentagon planners up at night, their worries compounded by recent high usage of limited precision weapons stocks. Beijing's other major problem is that, so far at least, Taiwan's always complex politics may not move in the direction that it wishes. It is far from clear whether KMT leader Cheng's embrace of Beijing will win over voters - and even if it does, her language of rapprochement still stops well short of calling for annexation by China. Nor is it clear that Beijing will do better in its efforts to isolate the new government in Tokyo, where for now it looks to be making little if any progress. European and Pacific nations' worries about future U.S. policy have, if anything, made many of them more enthusiastic than before about working with Japan or South Korea. Beijing has also yet to find its own way through the Gulf situation. Chinese-owned and -flagged oil tankers have periodically made it through Hormuz - but at nowhere near the pre-crisis rate. And the global supply chains on which China also depends are already looking rocky. Beijing may yet have to reach some kind of deal with Washington as it struggles to extricate itself from another Middle East war - perhaps even through the so-far deadlocked U.N. system - to pressure Iran to let more ships through. The Middle East crisis has without doubt created opportunities for China, but also delivered warnings for an ambitious coming superpower. (Writing by Peter Apps; Editing by Kevin Liffey)
Vercel Confirms Security Breach as Hacker Demands $2 Million and Claims to Sell Internal Access - Unchained
Vercel traced the breach to a compromised third-party AI tool and says sensitive environment variables were not accessed, but Web3 teams hosting on the platform are rotating credentials. Vercel, the cloud deployment platform that underpins frontend infrastructure for thousands of applications including many Web3 projects, confirmed a security breach on April 19 after a threat actor posted on BreachForums claiming to be selling stolen data for $2 million. The listing claimed to include access keys, source code, database content, and API tokens, including NPM and GitHub tokens tied to internal deployments and developer environments. Vercel said the intrusion originated from Context.ai, a third-party AI tool used by an employee, whose compromised Google Workspace connection allowed attackers to escalate access into Vercel's internal systems. CEO Guillermo Rauch confirmed this in a post on X. The company said environment variables marked as "sensitive" are stored in a way that prevents them from being read, and that there is no current evidence those variables were accessed. It has not disclosed how many customers were affected beyond describing the impact as limited to "a subset of customers," who are being contacted directly. For crypto and Web3 teams the breach is particularly sensitive. Vercel is the primary steward of Next.js, one of the most widely used web development frameworks, and hosts wallet interfaces, decentralized app dashboards, and front-end deployments for a significant portion of the industry. Projects storing API keys, private RPC endpoints, or backend service credentials in non-sensitive environment variables face potential exposure. Security researchers immediately advised all Vercel customers to rotate credentials and audit access logs for activity between April 17 and April 19.
Explainer-What Do We Know About Anthropic's Mythos Amid Rising Concerns?
April 20 (Reuters) - Anthropic earlier this month debuted Mythos, its most advanced AI model to date, equipped with sophisticated capabilities and designed for defensive cybersecurity tasks. Mythos' vast capabilities have sparked fears about the threat to traditional software security after the AI startup said the preview had uncovered "thousands" of major vulnerabilities in "every major operating system and web browser." HOW WAS THE MODEL LAUNCHED AND WHO HAD ACCESS TO IT? Anthropic has rolled out Claude Mythos Preview through a controlled initiative called "Project Glasswing", granting access to tech majors including Amazon, Microsoft, Nvidia and Apple. The company also extended access to a group of more than 40 additional organizations that build or maintain critical software infrastructure. WHAT ARE THE CONCERNS AROUND MYTHOS? Experts warned that the model can identify and exploit previously unknown vulnerabilities faster than companies can repair them. Its advanced coding and autonomous capabilities could dramatically accelerate sophisticated cyberattacks, particularly in sectors such as banking that rely on complex, interconnected and often decades-old technology systems, they have said. While debuting Mythos, Anthropic said the model's ability to find software flaws at scale could, if misused, pose serious risks to economies, public safety and national security. U.S. software stocks tumbled on April 9 after the Mythos launch on April 7 reignited fears that advances in AI could disrupt traditional firms. WHAT HAS THE WHITE HOUSE AND REGULATORS SAID ABOUT MYTHOS? The White House has held discussions with Anthropic CEO Dario Amodei about Mythos, with officials saying they talked about collaboration, cybersecurity and balancing AI innovation with safety. The talks were held despite the Pentagon slapping a formal supply-chain risk designation on Anthropic. The U.S. government is planning to make a version of Mythos available to major federal agencies, Bloomberg News has reported. Reuters reported that U.S. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell held a meeting with CEOs of major U.S. banks to brief them on the potential risks from the model. The model also raised alarm bells in Britain, with authorities holding talks with major banks and cybersecurity officials to assess possible risks. Banks are in close contact with their European regulators regarding Mythos, Christian Sewing, president of the German banking association and CEO of Deutsche Bank, said. (Reporting by Harshita Mary Varghese in Bengaluru)
Anthropic powers new Piraeus Bank AI hub
Greece's Piraeus Bank and Accenture have expanded their alliance with the launch of a dedicated AI Hub, powered by Anthropic. The initiative aims to speed up Piraeus' enterprise-wide AI transformation. The AI hub will become a central platform for designing, developing, and scaling advanced AI capabilities across the bank's entire value chain. The collaboration arrives amid heightened global scrutiny of Anthropic's latest model, Claude Mythos, which has been deemed "too dangerous" for public release due to its unprecedented ability to find and exploit software vulnerabilities. US and UK regulators have held talks with major banks over cybersecurity risks associated with the model and ensure they take measures to safeguard their networks. As for Piraeus, the new Anthropic-backed hub will be designed to reinvent banking processes in operations, customer experience, risk management, and compliance, while modernising its technology infrastructure. This will be facilitated through the integration of Accenture's industry and AI expertise - including its Data & AI Center of Excellence in Athens - with Piraeus' AI roadmap. Piraeus group chief operating officer Harry Margaritis called the hub a "strategic inflection point". Anthropic head of Southern Europe, Middle East & Africa Thomas Remy said: "Claude is built with the safety, reliability and transparency that highly regulated industries like banking demand. In partnering with Anthropic to power a new AI hub for Greek banking, Piraeus and Accenture have underscored our shared commitment to safe, responsible AI deployment." Besides, the hub will build Piraeus' long-term AI capabilities by developing, and upskilling specialist talent through targeted hiring and structured training programmes, including Udacity, Accenture's AI-native learning platform. This will help embed AI skills and new ways of working deeply across the organisation. A core priority of the partnership is the creation of secure, responsible, and human-centric AI solutions capable of autonomously supporting decision-making, streamlining complex processes. Piraeus and Accenture, through its newly formed Anthropic Business Group, will harness Anthropic's AI models and platforms to drive "innovation in a responsible manner", noted the professional services firm. The alliance ensures that all AI solutions align with the bank's values and regulatory requirements, enabling the development of secure and scalable AI applications. The new hub builds upon Piraeus' existing tie-up with Accenture to implement a cloud-first operating model. Meanwhile, in March, Commonwealth Bank of Australia (CBA) announced an expanded partnership with and investment in Anthropic.
The Polymarket Platform in Discussion for a $400 Million Fundraising
Polymarket is negotiating a new $400 million fundraising based on a valuation of about $15 billion. The prediction market is no longer just a corner of crypto; it is now a field that Wall Street is watching very closely. Polymarket is not just looking to raise new capital. The platform also wants to attract other strategic investors alongside Intercontinental Exchange, the parent company of the New York Stock Exchange. And this, even though the majority of users remain losers on these markets. In total, this round could reach up to $1 billion. This point is central. In March 2026, ICE already injected $600 million into Polymarket, as part of a broader commitment officially announced by the group. This support changes the perception of the case: we are no longer facing a marginal crypto startup, but an infrastructure that established players want to support. The targeted valuation, around $15 billion, remains lower than that of Kalshi, valued at around $22 billion during its last round. But this gap does not tell the whole story. It mainly shows that the sector battle is now played at a very high altitude, with valuations that would have seemed extravagant not long ago. Investor appetite does not come out of nowhere. Prediction markets exploded since the 2024 U.S. presidential election and have since maintained very high activity. TRM Labs mentioned in March a total monthly volume of more than $20 billion in January 2026, while several sector tracking show regular volumes well above $10 billion. It is no longer just a crypto fad. Nasdaq MRX filed in March to launch binary contracts linked to the Nasdaq-100. The move is revealing: traditional finance no longer looks at this market as a curiosity; it is already seeking its entry point. The same signal appears elsewhere. Charles Schwab has admitted to studying the subject, while Citadel Securities is also closely monitoring this segment. In other words, Polymarket may be raising money at the right time, just before the space becomes frankly crowded. At first glance, $15 billion for a betting platform on events may seem excessive. Yet, investors' logic is broader. They are not just paying for an interface where one bets on elections, sports, or economic announcements. They are paying for access to a new form of derivative product, simpler to understand and more viral to distribute.
OpenAI's top executive openly criticises company's biggest rival Anthropic in employee memo; says: 'Anthropic's story is built on fear, restriction and the idea that...
Denise Dresser, OpenAI's chief revenue officer, took direct aim at Anthropic in an internal memo sent to employees over the weekend -- calling out its rival's revenue accounting, compute strategy, and company philosophy in unusually blunt terms. The memo reads less like a strategic briefing and more like a case built against a courtroom opponent, dedicating significant space to dismantling Anthropic's narrative at a moment when the Claude-maker is gaining serious ground.The memo, viewed by The Verge, paints a picture of an OpenAI that is increasingly rattled by Anthropic's momentum. Dresser wrote that "the market is as competitive as I have ever seen it" and devoted a substantial chunk of the four-page document to picking apart Anthropic's business. The timing is notable: Anthropic's annualised revenue reportedly surged from $9 billion at the end of 2025 to $30 billion by March, fuelled largely by demand for its coding tools.Dresser's sharpest line was aimed squarely at Anthropic's identity. "Their story is built on fear, restriction, and the idea that a small group of elites should control AI," she wrote -- a characterisation that mirrors OpenAI CEO Sam Altman's long-running positioning of his company as the more democratising force in AI. In February, Altman had written that "Anthropic serves an expensive product to rich people."On the revenue dispute, Dresser accused Anthropic of using "accounting treatment that makes revenue look bigger than it is," specifically by grossing up revenue-sharing deals with Amazon and Google rather than reporting net figures. OpenAI claims the difference amounts to roughly $8 billion, which would drop Anthropic's stated $30 billion run rate to around $22 billion -- conveniently placing it behind OpenAI's reported $24 billion. Anthropic disputes this, saying it recognises gross revenue because it is the principal in those transactions, with cloud partners acting as distribution channels.Dresser also flagged what she described as a strategic blunder: Anthropic's conservative approach to acquiring compute. OpenAI says it has secured 8 gigawatts of computing capacity and is targeting 30 gigawatts by 2030, while it expects Anthropic to reach just 7 to 8 gigawatts by end of 2027. "Their strategic misstep to not acquire enough compute is showing up in the product," Dresser wrote, pointing to throttling and weaker availability as symptoms.Both companies are burning billions annually and racing toward IPOs, with investors now openly questioning OpenAI's $852 billion valuationThe memo's broader message was a rallying cry for OpenAI to out-execute on enterprise. Dresser urged the company to stop thinking of itself as a multi-product company and instead "think like a platform company with multiple entry points and one integrated enterprise offering." She emphasised that multi-year, nine-figure deals with enterprise clients are rising -- and that's the territory OpenAI now most wants to defend.Both companies are expected to go public this year, making the revenue and narrative war between them very much a preview of what investors will hear on the road to their respective IPOs.
