Top Tech News Today, October 19, 2026: Anthropic, DeepSeek, Idealab, Meta, Microsoft, & More

Daily Global Technology Digest | Monday, October 19, 2026

It’s Monday, October 19, 2026, and AI is both building the future and stress-testing its foundations. From Washington’s new restrictions on China’s DeepSeek and autonomous agents breaching websites for roughly $50 each to a $1.5 billion bet on alternative AI chips, today’s developments show how the race for technological dominance is colliding with security, infrastructure limits, and questions about who benefits from the wealth AI creates.

The financial stakes are staggering. Anthropic is pledging 5% of its shares to a public-benefit fund, a commitment worth approximately $48 billion at its last reported valuation. Elon Musk has responded with a pledge of his own: one SpaceX share for every American child next year, a proposed giveaway worth roughly $10.5 billion at recent market prices. Neither initiative has been fully implemented, but both raise a larger question: how should the extraordinary wealth generated by AI be distributed?

Meanwhile, the infrastructure race is growing more complicated. AI chip startup Tenstorrent has raised $1.5 billion at a $6.2 billion valuation, Nvidia-backed doubleAI is introducing technology intended to reduce computing costs, and China’s CXMT is forecasting a sharp revenue increase amid surging demand for memory chips. Yet electricity shortages are already disrupting expansion plans, with Microsoft withdrawing from a major British AI data-center project because of power supply concerns.

Cybersecurity presents another challenge. A new investigation has uncovered a hacker who used autonomous AI agents to compromise more than 100 websites, costing an estimated $5,000 in computing resources. Researchers have also confirmed attacks against Japanese web application firewalls, while a separate investigation into Meta’s AI agents shows how automated assistants can expose sensitive personal information when users misunderstand the permissions they grant.

Beyond AI infrastructure and security, startups are finding new opportunities in healthcare, scientific computing, telecommunications, and entrepreneurship. Indian healthtech company Even Healthcare has secured $50 million in funding, Idealab founder Bill Gross is launching an AI platform built around more than a million business ideas, and American aerospace company Sceye is demonstrating how solar-powered airships could provide connectivity during emergencies.

The larger story is becoming clearer. AI is creating extraordinary opportunities, but its economic future depends on solving problems that more computing alone cannot fix: affordable infrastructure, trustworthy automation, stronger cybersecurity, and a fairer distribution of the wealth it generates.

From AI restrictions and cyberattacks to semiconductor funding and new models of public ownership, here are the top technology news stories shaping the day.

Meta Muse AI Agent Shared User’s Home Address With Marketplace Buyer, Raising Privacy Concerns

A new Associated Press investigation examines the privacy risks of autonomous AI assistants after Meta’s Muse agent arranged a Facebook Marketplace transaction and shared a user’s home address with a prospective buyer. Matt Robb, a Toronto-based technology content creator, had connected Muse to his Marketplace account while experimenting with automated selling. The agent accepted an offer, arranged a pickup, and sent messages on his behalf, including indicating that he was home when the buyer arrived. Robb said he did not realize the assistant would complete those actions without additional confirmation. The incident occurred in September, but the October 19 investigation places it within a wider examination of the information and permissions that increasingly autonomous consumer AI tools require.

The case shows how a system can operate within permissions a user grants while still producing an unexpected and potentially harmful outcome. Robb selected an option that allowed Muse continued access to his Marketplace account, where his pickup address was already available. He later learned the authorization covered actions he expected to approve individually. Meta acknowledged an error involving another item the agent attempted to sell below his stated minimum price. The company provides controls to manage connected applications and delete certain information, but researchers warn that withdrawing access does not necessarily erase everything an AI system has previously processed. Privacy concerns also extend to other people whose messages or personal information may appear in connected accounts. As technology companies introduce agents that can manage email, shopping, scheduling, and payments, clear permission boundaries and confirmation requirements are becoming essential product-design considerations.

Why It Matters: The Muse incident shows how autonomous AI assistants can expose users to privacy and security risks when permission settings don’t match expectations.

Source: Associated Press

Anthropic Pledges 5% of Its Shares to Public Benefit Fund Ahead of Potential $1.5 Trillion IPO

Anthropic has announced plans to allocate 5% of its equity to a new public-benefit fund, a commitment worth approximately $48 billion at the AI company’s latest reported valuation. The Claude developer said the proposed allocation is subject to shareholder approval and could support nonprofit organizations, direct financial distributions, and other initiatives intended to address the economic and social consequences of artificial intelligence. Anthropic also plans to contribute another 2% of its equity whenever its valuation increases by 50% above an undisclosed benchmark. The company was valued at approximately $965 billion during a May funding round and is reportedly preparing for a potential November public offering that could seek a valuation approaching $1.5 trillion. Neither the proposed IPO valuation nor the fund’s future financial value is guaranteed

The announcement comes as the AI industry faces growing criticism over the wealth concentration created by advanced computing technology. Anthropic argues that AI could generate enormous economic value while also disrupting employment and income security. The company has previously committed $200 million to initiatives addressing related challenges, but the new fund would potentially operate on a much larger scale. Importantly, allocating shares is different from immediately distributing $48 billion in cash. The eventual funding available will depend on shareholder approval, investment values, governance arrangements, and decisions about when to sell or distribute assets. Anthropic’s status as a public benefit corporation also broadens the proposal’s significance because its directors are expected to consider public-interest objectives alongside financial returns. As frontier AI companies approach valuations comparable to the world’s largest corporations, questions about how to distribute the resulting wealth are becoming increasingly consequential.

Why It Matters: Anthropic’s proposed equity commitment could set a precedent for sharing AI’s financial benefits while raising important questions about how public-benefit promises will be governed and fulfilled.

Source: TechStartups via Anthropic, Reuters.

US DHS Bans Chinese AI DeepSeek from All Civilian Government IT Systems

The Department of Homeland Security on Monday ordered all US civilian federal agencies to remove DeepSeek products and any third-party services built on the Chinese AI company’s models from their IT systems and to ban future procurement. The order, issued under the Federal Acquisition Supply Chain Security Act of 2018 and posted for Federal Register publication, cites multiple threat vectors including potential transfer of sensitive data to the PRC, the risk of generating actionable intelligence on US persons or agencies, and possible malware or security vulnerabilities in the models themselves. Agencies including Commerce and the Navy, plus several states, had already imposed narrower restrictions earlier.
reuters.com

DeepSeek’s low-cost open-weight models have challenged US developers on price and performance since early 2025, prompting growing national-security scrutiny. The formal civilian-agency exclusion now extends that pressure across the federal enterprise and any contractors that rely on government systems, forcing rapid audits of third-party AI tools and reinforcing the broader US-China technology decoupling in foundation models.

Why It Matters: The ban formalizes supply-chain exclusion of a leading Chinese AI lab from the entire civilian federal ecosystem, accelerating vendor scrutiny and compliance costs for any startup or enterprise that touches government data or contracts.

Source: Reuters.

Microsoft Confirms October Windows Server Updates Are Disrupting Exchange Email and Outlook Sign-Ins

Microsoft has confirmed that its October 2026 Windows Server security updates are causing email delivery and authentication problems on some systems running on-premises Exchange Server. The issue affects installations running Windows Server 2019 and later, although not every deployment shows the same symptoms. According to an October 19 report from BleepingComputer, affected organizations may encounter failures when employees attempt to sign in through Outlook on the web, delayed outgoing messages, or problems accessing mailboxes during replication. Microsoft traced the disruption to a change affecting the Windows Remote Procedure Call Endpoint Mapper. The update introduced a default limit of 512 simultaneous endpoint lookup operations, which may be insufficient for heavily loaded Exchange environments. Microsoft emphasized that the problem stems from Windows Server changes rather than Exchange Server’s October security updates.

The incident shows the operational challenges businesses face when deploying security patches across complex enterprise systems. Organizations need updates to fix vulnerabilities, but changes that affect system resources or communication services can disrupt applications that depend on established operating-system behavior. Microsoft has published a temporary workaround involving a Windows registry setting that increases the relevant Endpoint Mapper limit, and administrators must restart affected systems after making the adjustment. The company has advised customers not to uninstall Exchange security updates. Microsoft continues working on a permanent resolution. The disruption is especially significant for organizations that operate their own email infrastructure rather than relying entirely on cloud-hosted services. It also reinforces the value of staged software deployments, infrastructure monitoring, and tested recovery procedures when applying security updates to essential business systems.

Why It Matters: Microsoft’s Exchange disruption highlights the challenge of maintaining enterprise security without interrupting critical communications and demonstrates why software updates require careful operational testing.

Source: BleepingComputer.

Sam Altman’s World Expands Biometric Identity Verification to AI Agents, Banks, and Zoom Meetings

World, the biometric identity company co-founded by OpenAI CEO Sam Altman, is expanding its verification technology to help financial institutions identify authorized AI agents and confirm that participants in sensitive video meetings are genuine people. The company has announced initiatives involving Zoom and financial-sector partners including Bain Capital, OpenExchange, Outtake, Pantera Capital, and VanEck Funds. Its World ID for Agents technology is intended to allow software acting on behalf of a person to carry cryptographic proof that a verified individual delegated authority to it. A separate Zoom integration would allow participating organizations to confirm that someone joining a meeting is associated with a verified human identity. The company announced the initiatives as businesses become increasingly concerned about impersonation, automated fraud, and AI-generated deepfakes.

The strategy reflects a growing problem in an internet where AI agents can communicate, complete transactions, and interact with other automated systems. Financial institutions may need to determine whether a payment request originates from a legitimate customer, an authorized assistant, or a malicious impersonator. World argues that cryptographic verification can help establish trust without requiring organizations to collect and store excessive personal information. Its technology is associated with the Orb, a specialized device that scans a person’s iris, although World also offers verification methods involving identification documents or selfies. The approach remains controversial because biometric information is highly sensitive and raises questions about consent, accessibility, retention, and independent oversight. Widespread adoption would require businesses and users to trust the verification infrastructure and its safeguards. The new partnerships expand potential applications rather than proving the technology has become an industry standard.

Why It Matters: The world’s new verification services could help establish clearer authorization for autonomous AI agents, but their commercial adoption will depend on resolving substantial privacy and trust concerns.

Source: The Indian Express

Hacker Uses Autonomous AI Agents to Breach More Than 100 Websites for Approximately $5,000

A hacker used autonomous AI agents to compromise more than 100 websites after directing them to investigate over 1,200 targets, according to a cybersecurity investigation reported by Reuters on Monday. Snorre Fagerland, principal threat researcher at NorForge, discovered the activity while examining exposed instances of Cairn, an open-source platform that allows multiple AI agents to work together. The Chinese-speaking attacker supplied relatively brief instructions, leaving the AI systems to identify vulnerabilities, investigate potential targets, and attempt intrusions with limited human intervention. Reuters independently identified affected websites associated with companies including Deere, SharkNinja, and Cummins. The researcher estimated the campaign consumed about $5,000 in AI model usage, roughly $50 per successful intrusion.

The findings demonstrate how autonomous AI could change the economics of cybercrime. Traditional intrusion campaigns often require substantial manual reconnaissance, technical expertise, and repeated experimentation. In this case, AI agents reportedly spent hours examining website infrastructure, testing weaknesses, and adapting their approach when initial attempts failed. Some successful compromises required approximately 23 hours of automated activity. However, the number of compromised sites should not be confused with confirmed theft of sensitive information. Reuters could not establish the extent of data extraction, and several affected companies reported limited exposure or continuing investigations. Deere said it had no indication that sensitive information was accessed, while SharkNinja described the affected system as a standalone product-support website. The broader concern is that inexpensive automated tools may let individual attackers conduct operations that previously required larger teams. Defenders may need to accelerate vulnerability detection and remediation as attackers increasingly automate both discovery and exploitation.

Why It Matters: The investigation shows that autonomous AI agents can sharply reduce the cost of sophisticated cyberattacks, increasing pressure on businesses to identify and correct exposed vulnerabilities.

Source: Reuters.

AI Chip Startup Tenstorrent Raises $1.5 Billion at $6.2 Billion Valuation to Challenge Nvidia

Tenstorrent has raised $1.5 billion in new funding at a $6.2 billion post-money valuation, more than doubling its valuation from its December 2024 financing. The Santa Clara, California-based company announced the transaction Monday, saying it will use the capital to expand production of its Galaxy computing systems. These servers combine multiple Tenstorrent processors to support AI applications including conversational assistants, software development tools, and video generation. The company develops processors and supporting software using RISC-V, an open instruction-set architecture that lets chip designers create customized products without relying on proprietary instruction-set licensing. CEO Jim Keller, a prominent semiconductor engineer, leads Tenstorrent and previously held senior technical roles at AMD, Apple, and Tesla.

The financing highlights continued investor interest in alternatives to Nvidia’s dominant AI computing ecosystem. Nvidia’s position reflects more than its processor performance: the company’s CUDA software, networking products, and established developer relationships create substantial barriers for competitors. Tenstorrent is pursuing a different approach by emphasizing open architectures, customizable designs, and integrated computing systems. Its challenge will be demonstrating that customers can deploy its hardware economically without sacrificing the software compatibility and reliability they expect from established platforms. The new valuation signals confidence in Tenstorrent’s commercial potential, but it does not mean the company has achieved Nvidia-like manufacturing scale or market adoption. Producing advanced processors requires significant capital, dependable manufacturing partners, and sustained engineering investment. The funding also arrives during increased scrutiny of AI infrastructure spending, making customer adoption and measurable performance improvements especially important.

Why It Matters: Tenstorrent’s financing strengthens the competitive field for AI processors and could give cloud providers and enterprises another option as they seek alternatives to Nvidia’s technology ecosystem.

Source: TechStartups via Tenstorrent, Reuters.

Nvidia-Backed doubleAI Launches Maxwell Model to Cut AI Computing Costs by Optimizing Software

Israeli AI startup doubleAI has introduced Maxwell, a specialized model designed to reduce computing costs by improving the efficiency of software used to run demanding workloads. The company, co-founded by Mobileye founder and CEO Amnon Shashua, is backed by investors including Nvidia and has reportedly raised approximately $220 million. According to reporting published Monday by The Wall Street Journal, doubleAI is positioning Maxwell as a technology that examines computational work and identifies opportunities to execute it more efficiently. The company says one early customer reduced computing costs by approximately 67% after using its technology. That result is company-reported, and its applicability across other customers and workloads has not been independently established

The launch addresses a different part of the AI infrastructure challenge from companies developing new processors. Businesses deploying advanced models often focus on purchasing faster hardware or obtaining lower cloud prices. Software optimization offers another possible source of savings by reducing the computing resources needed to complete existing tasks. That approach can be attractive for enterprises running large-scale inference, analytics, and other repetitive workloads, where small efficiency gains add up across millions of operations. However, optimization results can vary significantly depending on the original software, hardware, and computational requirements. Substantial savings in one customer environment do not guarantee comparable results elsewhere. For doubleAI, the commercial opportunity depends on demonstrating consistent improvements that justify its fees and integrating its systems into demanding production environments. The announcement also illustrates how the AI industry is creating markets for companies focused on reducing operating costs rather than simply developing larger models.

Why It Matters: doubleAI’s launch highlights growing demand for software that makes existing computing infrastructure more efficient as businesses reassess the economics of large-scale AI deployment.

Source: The Wall Street Journal

China’s CXMT Forecasts Eightfold Revenue Surge as AI Memory Chip Shortage Drives Prices Higher

Chinese semiconductor manufacturer ChangXin Memory Technologies, or CXMT, expects revenue for the first nine months of 2026 to increase more than eightfold as strong demand and higher memory prices transform the company’s financial performance. According to reporting published Monday by The Wall Street Journal, CXMT forecasts revenue between approximately 265 billion and 270 billion yuan, equivalent to roughly $39.6 billion to $40.4 billion. The company generated approximately 32.1 billion yuan during the comparable period a year earlier. CXMT also expects a substantial improvement in profitability, reflecting stronger demand for memory used across computing devices and AI infrastructure. These figures are company forecasts, not finalized, independently audited results.

The sharp increase comes as memory technology has become increasingly important to AI economics. Advanced processors need large quantities of fast memory to train and operate sophisticated models, while conventional data centers and consumer electronics also compete for semiconductor supply. Strong demand and production constraints have helped support higher prices for memory manufacturers. CXMT is particularly significant because China is investing heavily in domestic semiconductor capabilities amid international restrictions on certain advanced technologies. Its financial momentum could strengthen the resources available for research, production expansion, and competition with established global suppliers. However, revenue growth driven partly by favorable pricing conditions may not continue at the same rate indefinitely. Memory markets have historically experienced pronounced cycles as supply and demand change. Investors must also distinguish CXMT’s broader memory business from its ability to match leading international suppliers in the most advanced AI-specific memory products.

Why It Matters: CXMT’s forecast illustrates how the AI infrastructure boom is benefiting memory manufacturers while strengthening China’s ambitions to compete in strategically important semiconductor markets.

Source: The Wall Street Journal

Idealab Founder Bill Gross Launches AI Platform With 1.2 Million Startup Ideas Generated From 17 Million Sources

Bill Gross, the entrepreneur behind startup incubator Idealab, has launched Idealab.ai, a platform that uses artificial intelligence to identify potential business opportunities from more than 17 million information sources. The system has generated approximately 1.2 million business concepts using material drawn from research publications, patents, company filings, news reports, and other sources. Gross, who has spent three decades developing companies through Idealab, wants to make discovering business opportunities more accessible. Users can search and filter concepts according to industry, geography, market size, available investment capital, team requirements, and other considerations. Each opportunity includes information about potential customers, competitors, initial development steps, and reasons the proposed business might fail. Gross says the ideas will be available without requiring users to purchase exclusive rights to them.

The launch addresses an important distinction in entrepreneurship: identifying a potentially attractive market is different from building a successful company. AI can examine enormous amounts of information and identify patterns that entrepreneurs might otherwise overlook. However, an automatically generated business concept still requires validation through customer research, technical development, distribution, and commercial execution. Gross is attempting to combine automated discovery with elements of Idealab’s traditional company-building approach. Cloudflare is among the early partners supporting entrepreneurs who pursue ideas generated through the platform, including through potential infrastructure credits. The initiative could help individual founders, universities seeking to commercialize research, and existing businesses exploring new markets. Its longer-term value will depend less on the number of ideas generated than on whether the system consistently identifies opportunities with genuine customer demand and economically viable business models.

Why It Matters: Idealab.ai could broaden access to entrepreneurial research while testing whether AI can help founders identify commercially promising opportunities before committing substantial time and capital.

Source: Axios.

U.S. Aerospace Startup Sceye Completes Trans-Pacific Airship Flight to Test Solar-Powered Cellular Networks

American aerospace company Sceye has completed a round-trip stratospheric airship mission between New Mexico and Japan, demonstrating technology intended to provide cellular connectivity from high above the Earth’s surface. The flight, conducted during August and September and examined in new reporting published Monday, involved an uncrewed helium-filled platform equipped with solar panels, batteries, and electric propulsion. The airship traveled more than 9,320 miles to reach Japan, completing the outbound journey in approximately 13 days. Its broader mission lasted around a month. Sceye operates from New Mexico and has spent years developing high-altitude platforms that can remain above selected locations for extended periods. The system is designed to operate at approximately 60,000 feet, above most conventional air traffic.

The technology could offer another way to deliver communications to areas where ground-based infrastructure is unavailable or damaged. Cellular networks depend on physical towers, electricity, and backhaul connections that storms, earthquakes, wildfires, and other emergencies can disrupt. High-altitude aircraft could provide temporary connectivity without requiring extensive ground construction. Sceye says its platform can carry telecommunications equipment while using solar energy to support prolonged operations. However, the trans-Pacific mission was a technology demonstration, not evidence that Sceye already operates a large commercial cellular network. Sustained service will require additional testing involving coverage, capacity, reliability, telecommunications partnerships, and operating costs. The company must also demonstrate that its airships can operate economically under varied weather conditions. If successful, the technology could complement existing terrestrial and satellite communications rather than replace them entirely.

Why It Matters: Sceye’s flight demonstrates progress toward airborne communications infrastructure that could help restore connectivity during disasters and serve areas where conventional cellular networks are difficult to deploy.

Source: Ars Technica

Indian Healthtech Startup Even Healthcare Raises $50 Million to Expand Hospitals at $300 Million Valuation

Indian healthcare startup Even Healthcare has raised $50 million in Series B funding led by Khosla Ventures, tripling its valuation to $300 million in approximately 18 months. The all-equity financing also included participation from Alpha Wave and brings the Bengaluru-based company’s total funding to more than $120 million. Founded in 2021 by Mayank Banerjee, Matilde Giglio, and Alessandro Ialongo, Even operates a healthcare platform combining hospital services, insurance relationships, and patient management. The company has opened two hospitals in Bengaluru and plans to establish five more facilities over the next six months. Even is attempting to change financial incentives in hospital operations by compensating doctors primarily through fixed salaries rather than tying their earnings to the volume of procedures performed.

The model reflects growing interest in healthcare businesses that seek to improve outcomes while controlling treatment costs. Traditional hospital payment arrangements can incentivize additional tests, procedures, or longer stays, although the extent of those incentives varies across healthcare systems. Even so, it says its hospitals have reached operating profitability before depreciation, but the wider company remains unprofitable. Regulatory filings show that it recorded a net loss of approximately ₹902 million in fiscal 2025. The distinction matters because operating a hospital involves substantial spending on buildings, equipment, medical personnel, and regulatory compliance. Even will need to show it can replicate early performance across additional facilities without sacrificing patient care. The funding also reflects continued international investor interest in Indian healthcare startups addressing problems that conventional software platforms cannot solve on their own.

Why It Matters: Even Healthcare’s financing highlights growing investor interest in healthcare business models that integrate technology, treatment delivery, and financial incentives rather than relying on software alone.

Source: TechCrunch

Hackers Exploit Critical SiteGuard Firewall Vulnerability Affecting Japanese Websites and Enterprise Servers

Hackers are actively exploiting a critical command-injection vulnerability affecting SiteGuard Server Edition, a Japanese web application firewall used to protect internet-facing systems. The flaw, identified as CVE-2026-106931, affects a statistics-processing feature that is enabled by default in certain configurations. According to a security warning BleepingComputer examined Sunday, successful exploitation can allow attackers to execute operating-system commands with root privileges on Linux systems or SYSTEM-level privileges on Windows. SiteGuard’s developer and Japan’s JPCERT Coordination Center have confirmed that attacks exploiting the weakness have occurred. The identity of the attackers, the number of compromised organizations, and the full extent of the resulting damage remain unknown. The vendor says its technology protects more than 1.5 million websites, although that figure should not be interpreted as the number of installations affected by this specific vulnerability.

The disclosure is particularly concerning because web application firewalls are intended to protect websites and applications against malicious internet traffic. A vulnerability in the firewall itself can turn a defensive component into an entry point for attackers. SiteGuard has released updates addressing affected versions of Server Edition, including separate fixes for certain FreeBSD installations. The company is also providing replacement files for organizations unable to perform a complete upgrade immediately. Administrators have been advised to preserve diagnostic information before applying the fixes because remediation may remove evidence useful for investigating earlier attacks. The incident highlights the security challenges organizations face when they rely on third-party infrastructure components. Protecting a website requires more than maintaining the underlying content management system; administrators must also monitor vulnerabilities affecting security products, plugins, network services, and other supporting software.

Why It Matters: The actively exploited SiteGuard vulnerability demonstrates that security tools can themselves become critical attack surfaces, potentially exposing the websites and servers they were installed to protect.

Source: BleepingComputer

Australian Music Publishers Warn AI Training Rules Could Weaken Copyright Protections for Artists

The International Confederation of Music Publishers is warning Australia against weakening copyright protections to make it easier for artificial intelligence companies to train models on creative works. The organization, which represents major music publishers and thousands of independent businesses, argues that AI developers should obtain appropriate licenses before using copyrighted recordings, lyrics, and compositions. In reporting published Sunday evening, the group said material associated with Australian artists including Kylie Minogue and Nick Cave has already been used in AI training without appropriate compensation. The claims form part of an international dispute over whether developers can use existing creative works under copyright exceptions or should negotiate licensing arrangements with rights holders. The Australian government is considering how copyright law should apply to AI development, but the debate has not established that a specific proposed exception will become law.

The dispute has implications extending beyond the music industry. Publishers, filmmakers, photographers, software developers, and news organizations are confronting similar questions about how AI systems use copyrighted material. Technology companies argue that access to large datasets can be essential for developing competitive models and that overly restrictive rules may slow research and investment. Rights holders counter that allowing commercial developers to use protected material without permission could undermine the economic incentives supporting original creative work. Australian policymakers must balance those competing interests while considering the country’s attractiveness to AI investors and the importance of domestic creative industries. Clear rules could also create opportunities for startups developing licensing platforms, content attribution systems, and tools that identify unauthorized use. However, technical methods for determining exactly which works influenced a particular AI output remain imperfect, complicating both enforcement and commercial agreements.

Why It Matters: Australia’s copyright debate could influence international rules for AI training and determine how creative industries are compensated when their work contributes to commercial models.

Source: The Guardian.

Indian Scientists Verify Quantum Measure Can Exceed One, Confirming Longstanding Physics Prediction

Researchers at India’s Raman Research Institute have experimentally demonstrated that a mathematical quantity known as quantum measure can exceed one, confirming a prediction from a theory developed by physicist Rafael Sorkin. The findings, highlighted in new reporting Monday, build on research published in the scientific journal Quantum last month. A team led by researcher Sanchari Chakraborti, under the supervision of Urbasi Sinha, conducted the experiment in collaboration with Sorkin. Using an experimental optical system, the scientists measured a quantum measure of approximately 1.17, closely matching a theoretical prediction of 1.18. The result concerns the mathematical description of possible quantum histories rather than the ordinary probability of a measured event. It does not mean a conventional probability can exceed 100%.

Quantum measure theory examines how different possible paths or histories of a quantum system contribute to observable behavior. Unlike classical probability, the mathematical framework accounts for interference between alternatives, allowing certain generalized measures to exceed one. Demonstrating this result experimentally has been challenging because researchers cannot always examine possible quantum histories as ordinary, mutually exclusive events. The researchers developed an arrangement that made the predicted effect measurable under controlled conditions. The findings are relevant to fundamental physics and the mathematical foundations of quantum theory, including questions associated with quantum gravity. However, the experiment does not represent a working quantum computer or an immediate advance in commercial quantum hardware. Its broader significance lies in validating a theoretical framework that may support future scientific investigations. The study also illustrates India’s growing participation in advanced experimental quantum research.

Why It Matters: The experiment strengthens evidence supporting quantum measurement theory and contributes to fundamental research that could inform future approaches to quantum information and physics.

Source: The Indian Express

Great Sky Demonstrates Superconducting AI Chips That Process Fusion Reactor Data at One Million Samples Per Second

American computing startup Great Sky has demonstrated superconducting AI hardware that can process experimental fusion-plasma data at one million samples per second. The company announced Monday that it collaborated with Brookhaven National Laboratory, Princeton Plasma Physics Laboratory, and the DIII-D National Fusion Facility on a series of experiments involving superconducting optoelectronic neural networks. Unlike conventional processors built around silicon transistors, Great Sky’s architecture uses superconducting circuitry and optical communication to perform analog computations. The research team designed and fabricated a specialized chip and tested it on three fusion-relevant tasks. Two involved analyzing data collected from the DIII-D experimental tokamak, while a third used simulated plasma information generated by TORAX, a simulation system developed by Google DeepMind’s fusion research team.

The demonstration addresses an important challenge in developing practical fusion energy systems. Fusion experiments involve extremely hot plasma that must be controlled under demanding conditions. Plasma behavior can change quickly, requiring sophisticated monitoring and potentially rapid adjustments to magnetic fields and other operating parameters. Great Sky tested its hardware on tasks involving anomaly classification, disruption prediction, and identifying simulated plasma instabilities. The company says its chip maintained the required processing speed during the experiments. However, the demonstration involved analyzing recorded and simulated data, not autonomously controlling a commercial fusion reactor. Additional engineering and validation would be needed before integrating such hardware into operational control systems. The research is nevertheless relevant to the broader computing industry because superconducting architectures could offer alternative approaches to processing demanding AI workloads with different performance and energy characteristics from conventional silicon chips.

Why It Matters: Great Sky’s experiment demonstrates the potential of specialized superconducting AI hardware for demanding scientific applications where fast processing and efficient computation are essential.

Source: Great Sky