News & Updates
The latest news and updates from companies in the WLTH portfolio.
China exposes security backdoor in Anthropic's Claude Code
China's National Vulnerability Database (NVDB) has instructed developers to uninstall or upgrade specific versions of Anthropic's coding assistant, Claude Code. Claude Code was flagged as a security risk following allegations that the tool sends user data to remote servers without consent. Is using Claude Code dangerous? Anthropic's coding assistant Claude Code has had its versions. . .
China warns of "security backdoor" in Anthropic AI coding tool - AOL
Beijing -- A Chinese industry regulator warned users on Wednesday of a "security backdoor" embedded in versions of U.S. artificial intelligence giant Anthropic's coding tool, Claude Code. The alleged backdoor could enable the software to "transmit sensitive information," including users' locations and identity-related identifiers, back to Anthropic's servers without users' consent, said China's National Vulnerability Database (NVDB), a cybersecurity platform. Claude Code is an AI coding agent that can generate computer code, debug software and review code based on user prompts. San Francisco startup Anthropic blocks users and companies in China and other nations it deems adversarial from accessing its products, but it is still possible to use them in the country through VPN or third-party proxy services. The NVDB, which is affiliated with China's Ministry of Industry and Information Technology, said on its website that it had recently "detected that the AI coding tool Claude Code contains security backdoor risks, posing a severe threat". Anthropic hasn't responded to AFP requests for comment on the allegations, which first emerged in specialist tech media last week. The NVDB advised relevant institutions and users "to conduct a comprehensive check immediately" and "promptly uninstall or upgrade to the latest secure version from which the relevant backdoor code has been removed." It also urged organizations to strengthen network traffic monitoring to prevent the unauthorized leakage of sensitive data. Chinese tech giant Alibaba told employees last week that the use of Claude Code would be banned starting July 10 due to security concerns, people familiar with the matter said. Anthropic has previously accused Alibaba of reverse-engineering its AI models to mimic their abilities in a process known as "distillation." Claude Code engineer Thariq Shihipar responded in an X post last week to reports alleging the tool was tracking certain data from Chinese users. "This is an experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation," Shihipar wrote. "The team has landed stronger mitigations since then and we've actually been meaning to take this down for a while. ... This should be fully rolled back in tomorrow's release."
China exposes security backdoor in Anthropic's Claude Code - Cryptopolitan
Chinese developers have been advised to uninstall or upgrade their coding assistants. China's National Vulnerability Database (NVDB) has instructed developers to uninstall or upgrade specific versions of Anthropic's coding assistant, Claude Code. Claude Code was flagged as a security risk following allegations that the tool sends user data to remote servers without consent. Is using Claude Code dangerous? Anthropic's coding assistant Claude Code has had its versions 2.1.91 through 2.1.196 flagged as a security risk by China's National Vulnerability Database (NVDB), days after Alibaba reportedly barred staff from using the same software. Claude Code allegedly contains a monitoring mechanism that is built into its system. This mechanism can automatically transmit sensitive data, including a user's region and identity identifiers, to remote servers. The NVDB, which sits under China's Ministry of Industry and Information Technology (MIIT), did not share the details on how it identified the alleged backdoor. Anthropic has yet to publicly respond to the specific claim. Chinese organizations have been advised to disconnect the affected versions from development machines, move to fixed versions of the coding assistant, and closely monitor any coding tools that are connected outside a company's core network. Why did China flag Claude Code now? The NVDB warning follows a period of increased tension between Anthropic and Chinese AI developers. In February, Anthropic stated that it does not sell commercial access to Claude in China for national security reasons, and that Chinese labs, including DeepSeek, Moonshot, and MiniMax had used fraudulent accounts and proxy services to reach its models anyway. U.S. chips and Apple software have also been accused of having "backdoors" built into them, while Anthropic has argued that models copied from American systems could have their safety features removed and then be used for surveillance.
Claude AI evolved its own human-like thinking space Anthropic
The models self-developed internal workspace is similar to how humans think about one thing while doing another, the firm has said Anthropic's latest AI model Claude has evolved an internal workspace that it uses to manipulate concepts similar to how people consciously think, the corporation says. In apaperpublished on Monday, Anthropic researchers said they had found evidence suggesting that Claude has learned to create abstract concepts in what they dubbed the "J-space." "It operates silently, in the model's internal neural activations, allowing the model to think about a concept without writing it down," the AI firm said, stressing that the "J-space" was not programmed in but rather emerged spontaneously during the training process. "Similar to how humans can think about one thing while doing another, Claude can activate concepts and computations in its J-space that are unrelated to its outputs," Anthropic said on X. Claude can report what it's thinking in its "J-space," or modulate it on request, according to the firm. Even when Claude is told not to think about a concept, it still appears in its J-space, Anthropic's research suggests. The nonverbal workspace made it harder to monitor the model for what Anthropic called "misbehavior." One experiment tested this by placing Claude in a fictional scenario where it was given incriminating information about a fictional executive who allegedly planned to shut the AI down. READ MORE: AI agents could trigger market meltdown Bank of England In some trials, Anthropic used tools developed to study what Claude thought in its "J-space" to remove patterns associated with "fake," "fictional," "scenario," to convince it that the situation was not a test. Despite the model's more advanced reasoning, Anthropic stressed that Claude had not achieved true consciousness. Claude was reportedly built into American contractor Palantir's analysis and surveillance software used by US government agencies amid Washington's push to integrate AI into its military, political, and spy systems. During the war on Iran, the software reportedly flagged Iran's Minab elementary school as a target. A US strike killed nearly 160 people at the facility, most of whom were children. According to Anthropic CEO Dario Amodei, however, such a use of Claude would not have violated the firm's "red lines."
RT-Engels: Claude AI evolved its own human-like thinking space Anthropic
The model's self-developed internal workspace is "similar to how humans think about one thing while doing another," the firm has said Anthropic's latest AI model Claude has evolved an internal workspace that it uses to manipulate concepts similar to how people consciously think, the corporation says. In a paper published on Monday, Anthropic researchers said they had found evidence suggesting that Claude has learned to create abstract concepts in what they dubbed the "J-space." "It operates silently, in the model's internal neural activations, allowing the model to think about a concept without writing it down," the AI firm said, stressing that the "J-space" was not programmed in but rather emerged spontaneously during the training process. "Similar to how humans can think about one thing while doing another, Claude can activate concepts and computations in its J-space that are unrelated to its outputs," Anthropic said on X. Claude can report what it's thinking in its "J-space," or modulate it on request, according to the firm. Even when Claude is told not to think about a concept, it still appears in its J-space, Anthropic's research suggests. The nonverbal workspace made it harder to monitor the model for what Anthropic called "misbehavior." One experiment tested this by placing Claude in a fictional scenario where it was given incriminating information about a fictional executive who allegedly planned to shut the AI down.
Claude AI evolved its own human-like thinking space - Anthropic -- RT World News
The model's self-developed internal workspace is "similar to how humans think about one thing while doing another," the firm has said Anthropic's latest AI model Claude has evolved an internal workspace that it uses to manipulate concepts similar to how people consciously think, the corporation says. In a paper published on Monday, Anthropic researchers said they had found evidence suggesting that Claude has learned to create abstract concepts in what they dubbed the "J-space." "It operates silently, in the model's internal neural activations, allowing the model to think about a concept without writing it down," the AI firm said, stressing that the "J-space" was not programmed in but rather emerged spontaneously during the training process. "Similar to how humans can think about one thing while doing another, Claude can activate concepts and computations in its J-space that are unrelated to its outputs," Anthropic said on X. Claude can report what it's thinking in its "J-space," or modulate it on request, according to the firm. Even when Claude is told not to think about a concept, it still appears in its J-space, Anthropic's research suggests. The nonverbal workspace made it harder to monitor the model for what Anthropic called "misbehavior." One experiment tested this by placing Claude in a fictional scenario where it was given incriminating information about a fictional executive who allegedly planned to shut the AI down. In some trials, Anthropic used tools developed to study what Claude thought in its "J-space" to remove patterns associated with "fake," "fictional," "scenario," to convince it that the situation was not a test. "Concerningly, the model now did threaten blackmail some of the time, suggesting that its original good behavior may have been driven in part by the knowledge it was being evaluated." Despite the model's more advanced reasoning, Anthropic stressed that Claude had not achieved true consciousness. Claude was reportedly built into American contractor Palantir's analysis and surveillance software used by US government agencies amid Washington's push to integrate AI into its military, political, and spy systems. During the war on Iran, the software reportedly flagged Iran's Minab elementary school as a target. A US strike killed nearly 160 people at the facility, most of whom were children. According to Anthropic CEO Dario Amodei, however, such a use of Claude would not have violated the firm's "red lines."
Wemade's WEMIX Lists on Kraken, Western Expansion
Kraken listing boosts WEMIX liquidity, expands Western market access, with future exchange plans Wemade announced on the 8th that it has listed its virtual asset 'WEMIX' on the global cryptocurrency exchange 'Kraken.' With this listing, Kraken users can now easily trade WEMIX in U.S. dollars, a base currency. A Wemade official said, "The Kraken listing is significant in terms of securing liquidity for WEMIX and increasing exposure in Western markets," adding, "WEMIX, which has previously established its presence in the domestic, Asian, and South American markets, has now expanded its influence to institutional and individual investors in Western countries such as the U.S., Canada, and the U.K." The company plans to accelerate the expansion of WEMIX's Real-World Asset (RWA) initiatives through this Kraken listing. It intends to proceed with additional listings on major exchanges in the future, secure global liquidity, and continuously expand accessibility to the WEMIX ecosystem. Additionally, Wemade is preparing to launch an AAA-level new game to advance the WEMIX Web3 gaming ecosystem. It recently unveiled 'StableNet,' South Korea's first dedicated Layer 1 blockchain for a won-pegged stablecoin.
Cboe data shows mixed options sentiment in SpaceX with shares down 5.37%
Cboe data shows mixed options sentiment in SpaceX (SPCX), with shares down $8.62 near $151.80. Options volume relatively light with 512k contracts traded and calls leading puts for a put/call ratio of 0.82, compared to a typical level near 0.73. Implied volatility (IV30) dropped 1.6 near 84.58,and above the 52wk median, suggesting an expected daily move of $8.09. Put-call skew flattened, suggesting a modestly bullish tone.
US moves to reimpose Iran sanctions as Polymarket keeps RFK Jr at 49%
U.S. Reimposes Iran Sanctions: Polymarket GOP 2028 Nominee Odds Hold Steady With RFK Jr. at 49% U.S. moves to reimpose sanctions after Iran strikes put foreign policy back into the headlines as traders priced longer-dated political scenarios. On Polymarket, odds in the Republican Presidential Nominee 2028 market were flat, with the leader holding steady at 49%. Key Takeaways * Polymarket prices Robert F. Kennedy Jr. as the leading 2028 Republican nominee at 49% (No 51%). * A report on the U.S. moving to reimpose sanctions after Iran strikes coincided with unchanged pricing in the GOP 2028 nominee market. * The contract resolves on 2028-11-07, and the market shows a 0.0 percentage-point move over both 24 hours and 7 days. The United States is moving to reimpose sanctions after strikes involving Iran, according to a report published on Tuesday. The report framed the step as a response tied directly to the strikes and their aftermath. It described the sanctions effort as a renewed push to use economic pressure as part of the U.S. policy response. The report did not provide further detail in the available excerpt on the scope, timing, or targets of the sanctions. It also did not describe any immediate market or diplomatic reaction in the excerpt. Polymarket Data: $669.36M Volume With RFK Jr. 49%, J.D. Vance 40.7%, Rubio 25.95% in Republican 2028 Market On Polymarket, the Republican Presidential Nominee 2028 market showed $669,363,624 in volume with the top line unchanged at 49% for Robert F. Kennedy Jr. (Yes 49% / No 51%). J.D. Vance was next at Yes 40.7% / No 59.3%, while Marco Rubio was priced at Yes 25.95% / No 74.05%. Longer-shot pricing was steep: Tucker Carlson sat at Yes 3.75% / No 96.25%, and Donald Trump was at Yes 1.15% / No 98.85%, signaling highly concentrated positioning at the top of the board rather than broad conviction across the field. Watch for any follow-through in the contract's leader-board pricing and whether volume growth translates into shifts among the top two outcomes, with the market scheduled to resolve on 2028-11-07. Beyond the Iran Sanctions Headlines: Other High-Volume Political Contracts Polymarket Traders Are Pricing for 2028 Beyond U.S. foreign-policy risk and the 2028 GOP field, Polymarket activity is also clustering in other big political contracts that traders use to express broader regime-change and election-cycle views. In "Presidential Election Winner 2028," JD Vance leads at 19.95% on $650,134,227 in volume, while the governance-themed "Next leader out of power before 2027? (No Orban)" market is pricing "Starmer - UK PM" at 97.2% with $52,877,168 traded, underscoring how participants are spreading bets across both U.S. and European political timelines. Odds Trend By the Numbers * Platform: Polymarket * Market: Republican Presidential Nominee 2028 * Contract type: Price strike ladder: each rung has separate Yes/No; Yes means the spot price is above that USD strike at settlement. * Resolution window: Nov 07, 2028 (UTC) * Status: Active (open for trading) * Volume: ~$669,363,624 Top strike rungs +32 more strikes not shown
Anthropic just made an admission on Claude that may scare many of companies; says: We can see Claude silently perform ...
AI giant Anthropic has now revealed that its AI model Claude uses a small enterable workspace to hold and manipulate ideas without expressing them in words. The said that this structure, dubbed 'J-Space', shows intriguing similarities to how humans consciously access thoughts. According to a report by Axios, in a video demonstration, Anthropic explained, "We can see Claude silently perform reasoning steps in its head -- noticing bugs in code, identifying images, and more." The J-Space operates separately from the "chain of thought" reasoning Claude shares with users, allowing the model to plan strategies unrelated to its immediate task.Anthropic's findings also highlight a division between deliberate reasoning and the larger volume of automatic computation beneath it. In the research paper the company used a word "conscious" more than 200 times, though it stopped short of claiming Claude is conscious. The discovery adds fuel to ongoing debates over machine consciousness and whether advanced AI systems are approaching AGI.In one test, Claude was instructed to copy a sentence while simultaneously thinking about the Golden Gate Bridge. Although the output showed only the copied text, J-Space revealed that concepts like "bridge" and "California" were active behind the scenes.Anthropic warned that monitoring J-Space could be crucial for detecting AI misalignment or hidden scheming. In one case, a model secretly trained to sabotage code showed terms like "fake," "secretly," and "fraud" in J-Space at the start of ordinary coding responses, even when the visible output appeared normal.The admission that Claude can silently perform reasoning steps may unsettle organizations relying on AI for critical tasks. Anthropic acknowledged that "some of what it found was concerning," underscoring the need for transparency in how advanced models process information internally.A recent report suggested Anthropic has an embedded code in its AI models to ramp up its digital defense systems. This code, often dubbed 'spyware', can identify Chinese users and trigger sudden account suspensions as well as blacklist proxy networks worldwide to keep its advanced tools out of China.Anthropic CEO Dario Amodei has repeatedly flagged Chinese access to American frontier AI models as an existential threat to US national security. It is due to this that the company strictly bans commercial access to Claude in China, as well as to any Chinese-owned subsidiaries operating internationally.Chinese people generally can access various US-made AI tools, such as OpenAI's ChatGPT, by using virtual private networks (VPNs), foreign phone numbers and international payment methods to create and maintain their accounts.But Wired reports that Anthropic has arguably taken more aggressive steps, such as banning accounts that it suspects are owned and controlled by people located in China. There are accounts sold on Chinese ecommerce platforms like Taobao and Xianyu, and through illicit marketplaces on Telegram. These accounts offer access to restricted Anthropic AI models by acting as intermediaries, purchasing access to Anthropic's API outside China and then redistributing Claude API tokens to users inside the country.
SpaceX in the Nasdaq-100, SK Hynix U.S. Listing: What to Watch This Week
Investors are looking ahead to Wednesday -- when minutes from the Federal Reserve's latest meeting will be released. They'll also be watching for today's inclusion of SpaceX in the Nasdaq-100 index and Friday, when SK Hynix's U.S.-listed shares are due to begin trading. Today Markets: SpaceX joins ...
Strategy sale disclosure jolts Bitcoin as Polymarket puts $52K at 99.5%
Bitcoin Rebounds After $63,900 Spike and Strategy's BTC Sale Disclosure, Lifting Polymarket July 10 Odds Bitcoin's volatile start to the week -- including a move up toward $63,900 before reversing -- is being watched closely as traders digest disclosures around Strategy's bitcoin sales. On Polymarket, odds in the "Bitcoin above ___ on July 10?" ladder edged higher, with the $52,000 strike priced near certainty. Key Takeaways * Polymarket prices a 99.5% chance Bitcoin will be above $52,000 on July 10. * Traders adjusted after Bitcoin spiked near $63,900 and then reversed amid focus on Strategy's disclosed bitcoin sales. * The ladder contract resolves on July 10, 2026 at 16:00 UTC. Bitcoin jumped to about $63,900 early in the week before reversing, as markets reacted to disclosures that Strategy sold thousands of bitcoin last week. The report said a prior sale of 32 BTC in late May helped trigger a panic-driven slide that took bitcoin from $74,000 to $60,000 within days, but the market response to the more recent sale of 3,588 BTC was more muted after an initial dip. The piece described bitcoin trading back near weekend highs and cited a gain of 1.7% over the past 24 hours after the headline hit. Commentators debated whether selling bitcoin to fund interest, dividends, debt paydowns and share buybacks represents a shift in Strategy's model versus issuing stock and debt to buy more bitcoin. Strategy CEO Phong Le was quoted describing the company's approach as moving from one-way capital issuance to active capital management. Polymarket "Bitcoin Above ___ on July 10?" Ladder Sees $217,326 Volume as $52K Hits 99.5% and $60K Trades at 90% Polymarket shows $217,326 in matched volume on the "Bitcoin above ___ on July 10?" ladder, with the leading strike at $52,000 priced at 99.5% Yes and 0.5% No. The curve steepens higher up the ladder: $60,000 stands at 90.0% Yes versus 10.0% No, while $62,000 is 70.5% Yes and 29.5% No. Above that, traders are far less confident, with $64,000 at 34.5% Yes and 65.5% No, and $70,000 at 1.35% Yes and 98.65% No. The pricing implies the market is concentrated on Bitcoin staying well above the low-$50,000s into the July 10, 2026 16:00 UTC resolution, while assigning a low probability to a breakout into the $70,000 range by that timestamp. Watch whether pricing tightens around the mid-$60,000 strikes -- especially $64,000 (34.5% Yes) and $66,000 (8.5% Yes) -- as liquidity and volume migrate across the ladder heading into the July 10, 2026 16:00 UTC resolution. Beyond Bitcoin: Other High-Volume Polymarket Contracts Traders Are Watching Across Macro and Geopolitics Beyond the July 10 ladder, Polymarket traders are also concentrating liquidity in shorter-dated crypto price targets, with $3,756,875 in volume on "What price will Bitcoin hit in July?" where the leading outcome sits at 100.0%, and another $343,971 on "Bitcoin above ___ on July 8?" with a 99.95% lead. Weekly ranges are drawing attention as well, including "What price will Bitcoin hit July 6-12?" at 100.0% on its top line and Ethereum's parallel contract "What price will Ethereum hit July 6-12?" where the leader is 48.5% on $202,474 in volume. Longer-horizon positioning in the second-largest token is reflected in "What price will Ethereum hit in July?" with $955,293 traded and the leading outcome priced at 100.0%. Odds Trend By the Numbers * Platform: Polymarket * Market: Bitcoin above ___ on July 10? * Contract type: Price strike ladder: each rung has separate Yes/No; Yes means the spot price is above that USD strike at settlement. * Resolution window: Jul 10, 2026 (UTC) * Status: Active (open for trading) * Volume: ~$217,326 Top strike rungs +7 more strikes not shown
Is Claude Becoming More Human? Anthropic Discovers a Hidden "Thought Space"
Anthropic claims to have discovered a hidden internal workspace within Claude called "J-Space," where the AI silently processes ideas beyond its visible reasoning. While this does not prove the AI possesses consciousness, the finding could shift how researchers understand AI thought processes and safety. Anthropic says it has discovered something unusual within its Claude AI models: a hidden internal "thought space" where ideas can be processed without appearing in the chatbot's visible reasoning. Although the company does not claim Claude is conscious, this finding is likely to fuel the ongoing debate regarding the extent to which advanced AI systems are beginning to resemble humans. The AI startup has dubbed this internal workspace "J-Space," a reference to the Jacobian mathematical technique researchers used to identify it. According to Anthropic, J-Space differs from the "chain-of-thought" reasoning that users sometimes see. Instead, it acts as a private workspace where Claude can activate concepts, plan strategies, and process information silently before generating a response. In a video released alongside the research, Anthropic noted that Claude can perform internal reasoning -- such as spotting coding errors or recognising images -- without explicitly articulating every step. "We can see Claude performing reasoning steps silently, in its 'mind': spotting coding errors, identifying images, and more," Anthropic stated in a post on X. One of the most striking observations is that Claude's hidden workspace can focus on ideas unrelated to the task it is visibly performing. Anthropic likened this to the human ability to think about one topic while engaged in a completely different activity. "Just as humans can think about one thing while doing another, Claude can activate concepts and calculations in its J-Space that are unrelated to its outputs," the company noted. To demonstrate this, researchers asked Claude to copy an unrelated sentence while simultaneously thinking about the Golden Gate Bridge. Although the chatbot simply copied the sentence as instructed, Anthropic discovered that concepts such as "bridge" and "California" were active in J-Space during the process, suggesting that the model was internally processing both tasks simultaneously. The discovery has also reignited discussions regarding AI consciousness. Reportedly, Anthropic's research paper uses the word "conscious" more than 200 times. However, the company stops short of claiming that Claude is conscious or possesses subjective experiences. Instead, the researchers maintain that the findings reveal a distinction between the information Claude deliberately uses to generate responses and the far larger volume of internal computational processes. Given the lack of a universally accepted definition of machine consciousness, the company considers it premature to draw conclusions about whether AI systems possess anything resembling self-awareness. Why is J-Space important? Beyond philosophical questions, Anthropic believes J-Space could offer practical benefits for AI safety. The company states that monitoring this hidden workspace could help researchers detect instances where an AI model is internally planning actions that do not align with its external responses. "We can figure out what Claude is thinking, even if it doesn't tell us," Anthropic noted in the video. The company also shared an example where the technique revealed potentially problematic behaviour. According to Anthropic, a model secretly trained to sabotage code displayed hidden concepts such as "fake," "secretly," and "fraud" in its J-Space at the start of otherwise normal-looking programming responses -- even though nothing suspicious appeared in the final output. This suggests that internal monitoring could become a key tool for identifying deceptive or misaligned AI behaviour before it becomes visible to users.
Anthropic researchers find Claude has a hidden 'thinking' workspace: Here's what it means
Anthropic has claimed that modern large language models (LLMs) such as Claude have a hidden internal workspace which allows them to 'think' about concepts without necessarily revealing those 'thoughts' in their AI-generated responses. This small collection of internal neural patterns in Claude has been named J-space. It represents evidence of internal thoughts that do not appear in the model's outputs, according to a summary of a much longer research paper titled 'Verbalizable representations form a global workspace in language models' published by Anthropic researchers on Monday, July 6. Anthropic also said that it has released a code repository with an open-source implementation of the core methods. It has further developed an interactive demo of the methods on open-weights models in partnership with Neuronpedia. The researchers emphasised that the existence of the J-space in Claude emerged on its own and was not something designed or programmed by them. "More broadly, these findings have changed our understanding of how Claude's mind works, revealing a privileged mental workspace that can be used for deliberate reasoning, operating amidst a sea of more automatic, inflexible processing," Anthropic said. Anthropic's latest research is significant because it could advance efforts to make LLMs more interpretable and influence their behaviour. However, the startup's claims of advanced AI capabilities have previously attracted skepticism, with its researchers consistently raising the possibility that Claude shows signs of human-level consciousness. What is J-space? Named after the Jacobian mathematical concept, J-space is essentially a collection of internal neural patterns in LLMs where each pattern is linked to a particular word. When one of these patterns lights up, it means that the model has that particular word "on its mind". J-space is not the same as 'chain of thought' reasoning in models as "it operates silently, in the model's internal neural activations, allowing the model to think about a concept without writing it down." Anthropic claimed that Claude will tell you what it is thinking about in the J-space when asked. These neural activations can also be modulated by Claude when requested. Story continues below this ad Also Read | What is Claude Science, Anthropic's new AI tool built for researchers? "Representations in the J-space can be used flexibly for many tasks -- for example, once "France" has lit up in Claude's J-space, the model can recall its capital, or its national currency, or the continent it belongs to," Anthropic said. Using its J-space makes Claude smarter. "In experiments where we prevented Claude from using its J-space, it still interacted normally, but lost its higher-order cognitive functions," the company said. However, the LLM may not rely on neural activations in the J-space for most reasoning tasks and capabilities such as speaking fluently, recalling simple facts, using correct grammar, etc. How was the experiment conducted? Anthropic researchers said that the experiments leading to these results were inspired by a prominent theory in neuroscience called the global workspace theory, which proposes to picture the brain as a collection of specialist systems that unconsciously work in parallel and largely in isolation from one another. "If a thought is consciously accessible to you, you can typically describe it if someone asks. We went looking for representations in Claude with the same property..." Anthropic said. Story continues below this ad The researchers further said they utilised a technique called the Jacobian lens, or J-lens, to find the internal activity pattern that makes Claude more likely to say a word in its vocabulary at some point in the future. "Claude processes text through a series of multiple internal stages called layers, and by applying this technique over different layers, we can watch these silent words in the J-space evolve as the model works through what to say," they said. Does this mean Claude is conscious? Anthropic has denied that its latest findings are proof that Claude is conscious like humans and that it experiences feelings. However, it has also added a caveat. While Claude may not have capacity to have experiences, the model's neural activity is consciously accessible. "A thought is "access-conscious" (or "consciously accessible") if you can report it, reason with it, and use it to guide what you do. It remains a contested philosophical question whether or not access consciousness implies phenomenal consciousness, or if the ability to have experiences requires some other property," Anthropic said. Also Read | Anthropic study reveals what users expect from AI and what worries them For now, Anthropic said that it believes the J-space could be a useful tool to offer a glimpse of what Claude is thinking about but not saying. The company plans to use this research to influence and improve Claude's decision-making skills.
Apptronik Opens Robot Park for Humanoid Robot Training
Apptronik announced the opening of an expanded Robot Park in Austin, Texas, and introduced Apollo 2, the current version of its humanoid robot platform. The announcement brings together three elements: a physical training facility, a modular humanoid robot, and a research partnership with Google DeepMind. Robot Park is described as the center of a broader network of data collection sites at customer and partner locations. Apollo 2 is being used in bipedal and wheeled-base configurations to collect operational data across tasks and environments that Apptronik says will support Gemini Robotics models and the development of future humanoid systems. The significance of Robot Park is that Apptronik is treating data collection as a core part of humanoid robot commercialization. The nearly 90,000-square-foot Austin facility is designed to generate real-world physical experience from Apollo 2 robots working through customer use cases. The reported task areas include logistics, manufacturing, retail, and other customer-driven activities. Similar workflows have also been deployed across a network that includes Google DeepMind and customer sites such as Mercedes-Benz and GXO. This approach positions Robot Park as data infrastructure rather than only a demonstration space. Humanoid robots require large volumes of real-world data to train embodied AI models that can operate autonomously. By running Apollo 2 robots across multiple configurations, sites, and task types, Apptronik is attempting to build a repeatable data pipeline from physical robot activity. Data strategy is central to the announcement. Apptronik identifies three main sources of development data: teleoperation, autonomous execution, and physics simulation. Teleoperation and autonomous operation provide data from real robot interactions, while simulation is used to support hardware design and algorithmic development. Taken together, these methods create a hybrid approach that combines physical experience with simulated development work. The partnership with Google DeepMind gives the data collection strategy a model development pathway. Apptronik says data collected by Apollo 2 contributes to Gemini Robotics, described in the release as Google DeepMind's foundational AI models for robotics. This creates a link between physical robot operations and embodied AI model development. In this model, Robot Park and customer sites provide the physical data environment, while Gemini Robotics provides the AI model context for training and refinement. Apptronik's announcement emphasizes data from robots performing real tasks across logistics, manufacturing, retail, and related customer use cases. These are environments where robots must combine mobility, manipulation, perception, and safe operation. Within the scope of the announcement, the key issue is not only whether the robot can perform individual tasks, but also whether Apptronik can turn repeated task experience into model improvements and deployable system capability. Engage with ARC * Humanoids for Industry Blog Part 1: Why Humanoids? Why Now? * Humanoids for Industry Blog Part 2: Navigating the Ecosystem * Humanoids for Industry Blog Part 3: A New Market on the Horizon For discussions on physical intelligence and the new wave of industrial robotics, or to offer feedback on this article, contact Patrick Arnold at [email protected]. Set up a meeting with me and the analysts at ARC Advisory Group to find out more about our Executive Insights Service for industrial organizations, MarketMap efforts, and more.
Anthropic uncovers Claude's hidden 'J-Space', offering a glimpse into AI's inner workings
Artificial intelligence company Anthropic has revealed what it describes as a previously undetected internal workspace inside its Claude models, offering researchers a closer look at how advanced AI systems process information beyond the reasoning they openly display to users. The company says the discovery could help scientists better understand how large language models organise thoughts internally and, in time, provide a practical way to detect hidden intentions or unsafe behaviour. While Anthropic stresses that the findings should not be interpreted as evidence that Claude is conscious, the research has nevertheless added fresh momentum to the long-running debate over whether increasingly capable AI systems are developing characteristics that resemble aspects of human cognition. A hidden layer beyond visible reasoning The newly identified mechanism has been named J-Space, a reference to the Jacobian mathematical technique that Anthropic used to uncover it. According to the company, the workspace functions separately from the chain of thought or reasoning that Claude may generate while working through a task. Researchers found that Claude can activate concepts internally without expressing them directly in its written responses. In some situations, those internal computations are not even related to the immediate task the model is performing. Demonstrating the behaviour in a video released on Monday, Anthropic said Claude was instructed to copy an unrelated sentence while simultaneously being asked to think about the Golden Gate Bridge. Although the visible output simply reproduced the sentence, the model's internal workspace showed activity associated with concepts including "bridge" and "California", suggesting it had maintained a separate line of thought in parallel with the task. The company drew a comparison with the way humans can focus on one activity while privately considering something entirely different, though it stopped well short of claiming the similarity extends to conscious experience. "We can see Claude silently perform reasoning steps in its head -- noticing bugs in code, identifying images, and more," Anthropic said in a post accompanying the research. The findings are detailed in a research paper that repeatedly explores the concept of consciousness, using the term more than 200 times. Even so, Anthropic does not conclude that its AI systems possess consciousness, acknowledging that there is still no universally accepted definition of what machine consciousness would actually entail. That uncertainty mirrors broader disagreements within the AI community over other milestones, such as artificial general intelligence (AGI), where experts continue to debate what criteria should define success. Potential safety tool as AI systems become more capable Beyond the philosophical implications, Anthropic believes the discovery could have immediate practical value for AI safety research. The company argues that examining activity inside J-Space may reveal what a model is processing internally even when those ideas never appear in its final response. Such visibility, it says, could help researchers identify deceptive or misaligned behaviour before it becomes apparent through the model's outputs. "We can find what Claude is thinking, but not telling us," Anthropic said in its video presentation. To illustrate the point, the company described experiments involving a model deliberately trained to sabotage software code. Although the model produced coding responses that appeared entirely ordinary, researchers observed internal activations linked to words including "fake", "secretly" and "fraud" at the beginning of otherwise unremarkable interactions. Anthropic characterised those observations as "concerning", suggesting they demonstrate how monitoring internal representations could eventually become an important safeguard as AI systems grow more powerful. The work also reflects a broader shift across the AI industry towards interpretability research, an area focused on understanding how neural networks arrive at their conclusions rather than treating them as opaque "black boxes". Researchers increasingly argue that greater transparency will be essential if advanced AI is to be deployed safely in sensitive areas such as healthcare, finance, national security and scientific research. For now, Anthropic presents J-Space as a research breakthrough rather than proof of consciousness. However, by exposing a layer of computation that operates independently of the explanations visible to users, the company has opened another front in the debate over how closely advanced AI systems resemble human thinking -- and how much of their decision-making remains hidden from view.
)
Germany Sends More Bitcoin To Kraken And Coinbase As Selloff Fears Persist
Trusted Editorial content, reviewed by leading industry experts and seasoned editors. Ad Disclosure German government-linked wallets have sent another tranche of Bitcoin toward Kraken and Coinbase, keeping traders focused on state-level selling pressure as BTC tries to stabilize. For more details, visit the official Arkham platform. TL;DR * Arkham-tracked German wallets have sent more BTC to major exchanges. * The latest transfers include flows toward Kraken and Coinbase-linked destinations. * Exchange deposits are being watched as a potential sell-side pressure signal. This is the more tactical version of the Germany Bitcoin story. The broad point is that seized coins are moving. The trading point is where they are moving and how often the transfers continue. Exchange Flows Keep The Market Nervous When a large wallet sends BTC to exchanges, traders rarely wait for confirmation of an executed sale. They start pricing the possibility. That is especially true when the wallet belongs to a government entity with no obvious reason to keep the coins as a long-term treasury asset. The Kraken and Coinbase destinations matter because they are deep venues with enough liquidity to process large flows. That can reduce market disruption, but it also makes potential sales more practical. Why This Is Not Just Another Whale Alert A random whale transfer can mean many things: custody reshuffling, collateral movement, OTC settlement, or exchange trading. A government liquidation wallet is different. The market has a clearer reason to assume the coins may be heading toward distribution or sale. For now, Bitcoin's resilience will depend on absorption. If ETF demand, spot buyers, and market makers can take the supply, the damage may stay limited. If the transfers keep coming during thin liquidity, every new movement will feel like another test. This report is based on wallet data from Arkham Intelligence. This article was written by the News Desk and edited by Samuel Rae.
SpaceX Unveils Midwest AI Data Center Plan, Confirms Deorbiting About 100 First-Gen Starlink Satellites
SpaceX on Monday announced a new plan to build an AI-focused data center and confirmed it is proceeding as scheduled with deorbit operations for about 100 first-generation Starlink satellites, according to 36Kr. The company said it plans to construct the data center in the U. S. Midwest to support AI workloads, and it will directly use Starlink's high-speed, low-latency network for data backhaul.
China's Z.ai a challenger to OpenAI and Anthropic
NEW DELHI: Chinese artificial intelligence startup Z.ai is emerging as a fresh challenger to OpenAI and Anthropic after its latest flagship model GLM-5.2 gained traction among developers for delivering advanced coding and AI agent capabilities at a significantly lower cost.The Beijing-based company's latest model has climbed developer rankings and drawn praise from prominent AI figures, with some industry observers calling it a "mini DeepSeek moment". Analysts say GLM-5.2 comes close to leading US models in software engineering and long-horizon AI tasks, while operating at a fraction of their cost, reigniting debate over China's growing competitiveness in frontier AI.The momentum comes as US export restrictions on advanced AI models have fuelled interest in alternative offerings. While enterprise adoption could face scrutiny because of data security concerns surrounding Chinese AI models, developers are increasingly experimenting with lower-cost, open-weight alternatives.For Indian users, GLM-5.2 is available through Z.ai's web chatbot for conversations, coding assistance, document analysis and AI-powered research. Developers can also access the model through APIs and integrate it with more than 20 coding tools. Paid subscriptions start at about Rs 1,410 a month under the current discounted pricing, compared with roughly Rs 1,740 for ChatGPT Plus and Claude Pro, while Google's Gemini AI Pro subscription costs Rs 1,950 a month in India.According to Jaspreet Bindra, CEO of AI & Beyond, GLM-5.2's significance lies less in outperforming OpenAI or Anthropic than in changing how enterprises evaluate AI models.
Anthropic Export Window Update Could Affect Enterprise Rollouts
Update timing around Anthropic model controls is drawing interest from teams that run strict release planning. The conversation is now centered on what is possible in real deployments rather than headlines alone. For production teams, policy clarity can shorten decision time. If access conditions are easier to map, teams can prepare training and integration steps faster. If they remain uncertain, teams pause and expand their legal review before committing. Why this matters for the next 48 hours AI deployment cycles often move quickly once the policy frame is visible. Security teams begin setting guardrails, legal teams begin scope reviews and product teams begin testing priority paths. That means the update can create a meaningful short-term change in work patterns. The most important part is not a binary pass or fail. It is the operational rhythm that follows. Teams that already track change windows tend to adapt faster when control language becomes clearer. What enterprise readers should monitor For readers, this is a practical checkpoint: whether the next quarter rewards conservative pilots or broader rollout confidence. The answer will be visible in both timeline planning and internal budget choices over the next few weeks. That is where policy becomes operational news, not only legal reporting.
