Naver began developing an artificial intelligence (AI) model specialized for cybersecurity after export controls on Anthropic’s Claude disrupted access to the service.

The disruption led the company to conclude that it could not rely on overseas models for its security capabilities.

Kim Jin-hwi, an executive director at Naver Cloud’s Security Division, disclosed the circumstances at the AIDC Security Technology Workshop held at Militopia Hotel in Seongnam, Gyeonggi Province, on Sept. 7. Kim, a former National Intelligence Service official, joined Naver in April. He oversees infrastructure security across the group, including Naver, Naver Cloud and Naver Webtoon.

“The service was interrupted midway because of export controls,” Kim said. “After experiencing that problem, we decided that we needed to build our own AI model.”

The US Department of Commerce in June ordered Anthropic to suspend access by all foreign users to Claude Fable 5 and Claude Mythos 5, citing national security concerns. Access to the two models was blocked for 19 days until the department lifted the restrictions. Mythos 5 remains available only to certain approved organizations.

Kim also pointed to limitations of overseas models. “Foreign models don’t have an understanding of the domestic threat environment,” he said. “Their guardrails, or AI control mechanisms, also pose major problems for how we use them.” He added that the episode made Naver realize it could not depend on overseas providers for its security policies and capabilities.

Kim said, however, that Naver does not plan to handle all AI security tasks with its own model. The company intends to continue using external solutions for tasks including vulnerability assessments.

According to Kim, Naver is already using existing AI models in actual security operations. AI automatically scans all ports on about 100,000 externally exposed servers twice a day. The company also operates a system in which a red team responsible for attacks and a blue team responsible for defense and patching exchange results and repeatedly validate vulnerabilities.

As Naver prepared its own model, it also built an in-house “harness” that connects AI models with vulnerability-analysis tools and execution environments. “Using this harness, we reproduced all of the open-source vulnerabilities that Mythos had discovered,” Kim said. Naver’s security team identified and reported 29 previously undisclosed high-risk vulnerabilities during the process, he said.

Against that backdrop, the government announced a project in late July to develop an AI foundation model specialized for cybersecurity. Naver Cloud formed a consortium and applied. “There were many areas that were difficult to address at the level of a single company,” Kim said. “It is an opportunity to build GPU resources, data and validation capabilities on a national scale.”

The consortium, comprising Naver Cloud and 32 other organizations, was selected on Sept. 3 as the final contractor for the government’s project to develop an AI foundation model specialized for cybersecurity.

Naver will contribute 4,000 of its own GPUs. The consortium aims to train Naver’s HyperCLOVA X and LG’s EXAONE on security data to develop a mixture-of-experts (MoE) security model with 700 billion parameters. The completed model will also be deployed on domestic AI chips from FuriosaAI and HyperAccel to validate its performance.

Development will begin on Sept. 18. With Minister of Science and ICT Bae Kyung-hoon saying the security-specialized model will be unveiled within the year, the consortium is targeting a demonstration in December. An interim evaluation will follow in February next year, with the project scheduled to conclude in July. Naver plans to release the final model as open source and distribute it through application programming interfaces (APIs) and software development kits (SDKs), allowing Korea’s cybersecurity industry to use it.

Smilegate's new title "Eclipse: The Awakening," or "Eclipse," launched on PC and mobile at noon on Sept. 10. The game is a key project for Smilegate after its operating profit fell 30% last year, while developer NPIXEL is seeking to offset an operating loss of 12.3 billion won. With Com2uS's new title "Zeus," released about two weeks ago, holding the top spot in mobile game sales, attention is turning to how Eclipse will perform.

Eclipse is a free-to-play game with in-app purchases. The company designed its monetization model to attract users early on while reducing the burden of spending. According to Smilegate, the game secured 1 million pre-registrations before launch. During the first round of character pre-creation on Aug. 19, all 24 available servers reached capacity within 25 minutes. Some streamer servers filled up in just one minute. Smilegate increased capacity and opened a second round the following day, but that also closed within a day.

The game's core concept is "MMOLite," designed to reduce the time commitment, spending pressure and intensity of player competition typically associated with massively multiplayer online role-playing games (MMORPGs). A key example is "Sanctuary," which accumulates character-growth resources and summon tickets even while users are offline. An AI mode allows gameplay to continue automatically after users exit the game. Player-versus-player (PvP) combat is also disabled in most areas.

The monetization system follows the same approach. Eclipse does not allow users to directly purchase "cash accessories" that boost character stats. Some growth-related products are subject to weekly or monthly purchase limits rather than being available without restriction. The goal is to minimize the gap between hardcore and casual players. From a financial perspective, however, the policy could lower average revenue per user (ARPU) by reducing the incentive to spend. The model therefore depends on maximizing the overall number of paying users.

The importance of Eclipse's commercial performance is evident in the financial statements of Smilegate and NPIXEL for last year.

Smilegate posted consolidated revenue of 1.4365 trillion won and operating profit of 359.8 billion won in 2025. Revenue fell 5.6% from the previous year, while operating profit declined 30%. Net income dropped about 36% to 302.3 billion won.

The stakes are even higher for developer NPIXEL. According to its audit report, the company generated 13 billion won in revenue last year. Sales have continued to decline from 67 billion won in 2022 to 31.6 billion won in 2023 and 14.2 billion won in 2024. NPIXEL posted an operating loss of 12.3 billion won last year. It cut operating expenses to about 25.4 billion won from 67.6 billion won a year earlier, narrowing its operating loss by about 77% from 53.4 billion won in 2024. Despite the cost reductions, revenue has yet to recover.

NPIXEL also faces pressure from its financial structure. At the end of last year, the company had 7.7 billion won in current assets, compared with 56.3 billion won in current liabilities, more than seven times as much. Total assets stood at 14.8 billion won and total liabilities at 66 billion won. Total equity was negative 51.1 billion won, leaving the company in a state of complete capital impairment. EY Han Young said in its audit report that material uncertainty exists that may cast significant doubt on the company's ability to continue as a going concern. NPIXEL's co-chief executives have also lent money to the company. Borrowings related to co-CEO Jung Hyun-ho totaled 11.9 billion won, while those related to co-CEO Bae Bong-gun amounted to 6.9 billion won.

Eclipse therefore carries significant weight for both companies. For Smilegate, it represents a new revenue source that could help revive earnings. For NPIXEL, it is a game that needs to rebuild the company's revenue base and help it move out of losses and capital impairment.

The game industry is focusing less on Eclipse's initial user numbers than on its future retention rate and sales ranking. Users drawn in by free summon tickets and other incentives need to remain in the game for at least a month, with some converting into paying customers. Smilegate's "MMOLite" model will only prove viable if the game can generate meaningful revenue while keeping the burden of spending low.

As of 4 p.m. on Sept. 10, Eclipse ranked No. 10 in sales on Apple's App Store. It ranked No. 1 in free downloads.

Zscaler has unveiled Agentic SOC, a security operations platform that uses frontier artificial intelligence (AI) models from companies including OpenAI and Anthropic.

The U.S. cloud security company said on Sept. 9 that AI-powered attacks are occurring at speeds that traditional security operations centers (SOCs) struggle to handle, requiring SOCs to be redesigned so AI can take the lead in detecting, investigating and blocking threats.

Zscaler's Agentic SOC combines security information and event management (SIEM), which collects logs and detects threats, with security orchestration, automation and response (SOAR), which handles identified threats. It adds AI agents that can independently assess situations and take action throughout the process.

The company has also partnered with frontier AI labs including Anthropic and OpenAI, based on its view that a single model or approach has limitations in delivering accurate agentic AI operations.

Agentic SOC assigns AI agents to perform decisions and responses traditionally handled by security analysts. The agents identify genuine threats among large volumes of security alerts, investigate attack paths and root causes, and determine whether a threat is present. They can then automatically take necessary actions, such as isolating compromised users or blocking attackers' command-and-control (C2) communications.

Zscaler provides the AI agents with security data collected from networks in real time and its own threat intelligence. The platform can also integrate with customers' existing security tools to analyze threat information and automatically block lateral movement by attackers across systems.

"Agentic SOC represents a fundamental reimagining of security operations," said Deepen Desai, Zscaler's vice president of cybersecurity. "It is designed to proactively reduce exposure and contain threats at machine speed."

"AI attacks operate with a speed, scale and adaptability that are markedly different from humans, fundamentally changing the threat landscape," said Allie Mellen, a principal analyst at Forrester Research. "Defenders need to focus on fundamentals such as zero-trust principles, data loss prevention and access restrictions."

STRADVISION is expanding its artificial intelligence (AI)-based vision perception technology, proven in mass-produced vehicles, into robotics and autonomous devices.

STRADVISION said Sept. 8 that it will join semiconductor design company Arm’s “Total Design for Physical AI” ecosystem. The announcement was made at the Arm Everywhere event in Shanghai, China.

STRADVISION plans to contribute vision perception technology accumulated over more than a decade in automotive mass-production environments to Arm’s physical AI ecosystem. Physical AI refers to AI technologies that perceive and interpret their surroundings and act in real-world environments, such as robots and autonomous vehicles.

STRADVISION has applied technologies including object detection and classification, depth estimation and semantic segmentation to front-facing cameras, surround-view systems and multi-camera systems. The company plans to extend its experience implementing AI vision technology in production vehicles under constraints on computing performance, power consumption, heat and cost to robots and other autonomous devices.

STRADVISION will also participate in developing Arm’s Robotics Capability Framework. The framework is designed to establish common standards defining what robots should perceive in different environments and the level at which they should independently make decisions and take actions.

Arm sees the cost and time required for system integration as key factors determining the pace of commercialization because robotic systems must combine technologies including semiconductors, AI models, middleware, development tools and applications. The common framework is intended to make it easier for companies using different semiconductors and software to design, compare and integrate robotic systems.

“The next challenge for physical AI is integrating increasingly advanced technologies into complete systems that can be deployed in real-world environments,” said Dermot O’Driscoll, vice president of Physical AI Business Development at Arm. “STRADVISION will bring proven expertise in perception technology.”

“For more than a decade, we have deployed perception technology in production vehicles across major automotive computing platforms,” said Philip Vidal, chief business officer (CBO) at STRADVISION. “We will support Arm’s efforts to create a common language for robotics.”

KT has signed a memorandum of understanding (MOU) with the Korea Advanced Institute of Science and Technology (KAIST) and physical artificial intelligence (AI) specialist DAIM Research to develop, demonstrate and commercialize physical AI technologies. The agreement follows discussions among the three organizations on continued business cooperation during the second phase of the Physical AI Alliance led by the Ministry of Science and ICT.

KT, KAIST and DAIM Research will cooperate on commercializing a dark factory platform for unmanned manufacturing, identifying physical AI business opportunities and expanding K-Physical AI into global markets.

Commercialization of the dark factory platform will center on the Ministry of Science and ICT's 2026 project to establish a research and development ecosystem for collaborative-intelligence physical AI-based software platforms. The project aims to develop and demonstrate AI-based unmanned and automated platforms that enable manufacturing sites to operate autonomously. The three organizations will participate in the project for five years.

The three organizations will focus on developing technologies to automate manufacturing processes based on logistics and manufacturing robot control platforms and robotics technologies held by KAIST and DAIM Research. They will also build a testbed similar to actual industrial environments to validate the technologies and expand their application to various manufacturing sectors based on the results.

KT will provide an operating platform that can integrate and manage various robots and AI systems based on its data platform. It also plans to support real-time decision-making and robot control using edge AI, which processes data close to robots to minimize latency, as well as high-speed, ultra-low-latency network infrastructure.

"The true competitiveness of physical AI does not end with developing good technology, but lies in making robots and AI work in actual manufacturing environments and validating their performance," said Choi Sung-yool, Senior Vice President for Research at KAIST.

Samsung SDS will deploy about 200 artificial intelligence (AI) specialists to corporate workplaces as part of efforts to strengthen its artificial intelligence transformation (AX) business. The company plans to expand its scope from analyzing and redesigning business processes to building and operating AI agents. It will also pursue an AI security business in cooperation with OpenAI.

Samsung SDS unveiled its "Multi-Dimensional AI Full Stack" strategy at REAL Summit 2026 in Seoul on Sept. 8.

"AI changes the way people work, but AX changes the business," Samsung SDS President and CEO Lee Junehee said. "AX is not simply about introducing AI to save time and costs, but about redesigning the business itself and the way work is done from the ground up."

One of the company's initiatives is the Forward Deployed Engineer (FDE), an on-site AI specialist. FDEs work directly at corporate workplaces to analyze real-world problems and design and implement AI-based solutions. The FDE model was created by Palantir for its enterprise software business 22 years ago.

Samsung SDS plans to expand its team of AX-focused FDEs to about 200 by the end of this year through cooperation with OpenAI and Anthropic. They will work on-site, handling everything from analyzing customer workflows to developing AI agents and deploying and operating systems. Unlike the conventional approach of installing pre-built systems, the model continuously modifies and applies AI based on a company's operational data. Samsung SDS introduced FDEs in July as a new model for enterprise AI adoption.

Samsung SDS also announced a "Client Zero" policy to transform its own operations with AI first. The company will apply AX to seven major internal functions — development, procurement, quality, corporate support, project execution, marketing and sales — and offer externally the methods it validates internally. In effect, Samsung SDS intends to go through the trial-and-error process before its customers do.

The company also announced plans to enter the AI security business. Samsung SDS will participate as a pilot partner in OpenAI's Daybreak Partner Program to develop cybersecurity services using AI models. AI will automatically identify and validate corporate security vulnerabilities and provide patching solutions. As with FDEs, security specialists will be deployed to customer sites to implement the services.

Samsung SDS also plans to launch a "robot orchestration platform" in 2027 to integrate and manage multiple types of robots. The initial target will be about 430 existing manufacturing automation customers. The platform is intended to expand the company's physical AI business combining robotics and AI and target the autonomous manufacturing market.

Taken together, Samsung SDS's AX strategy significantly broadens its business scope, from supplying AI infrastructure and applying AI to enterprise operations to AI security and robot operations at manufacturing sites.

Securities analysts are also watching whether Samsung SDS's AX expansion will translate into higher managed service provider (MSP) and AI revenue. Hyundai Motor Securities analyst Choi Seung-ho said in a Sept. 2 report that one of the key factors to watch is whether accelerating AX adoption leads to faster MSP revenue growth, setting a target price of 300,000 won. Choi forecast Samsung SDS's 2026 revenue at 14.3457 trillion won and operating profit at 810.3 billion won.

LS Securities analyst Sun Yujin on Aug. 19 described Samsung SDS as a pioneer in South Korea's IT services industry at the forefront of AX. Sun cited growth in AI revenue from Samsung Group's AI transformation and the expansion of intelligent work management platforms for the government as key growth drivers for the second half and beyond, raising the target price to 307,000 won.

Samsung SDS's second-quarter IT services revenue rose 5% from a year earlier to 1.7625 trillion won. Cloud revenue increased 17% to 779.4 billion won. Revenue from the MSP business, which includes AX projects, was 346.7 billion won, up 17% from a year earlier, driven by AX projects in the financial sector and increased enterprise resource planning (ERP) projects in the shipbuilding industry, the company said.

Inswave is stepping up its push into the Japanese market with artificial intelligence (AI)-based development technologies.

Inswave said Sept. 8 that 32 representatives from major Japanese information technology (IT) companies visited its headquarters in Yeouido, Seoul, on Sept. 4 to review the company’s AI and digital transformation (DX) strategy and key products and discuss potential business cooperation. The delegation included executives and engineers from Japanese IT companies in the energy, public and manufacturing sectors.

Inswave demonstrated its flagship solution, WebSquare AI. The platform uses generative AI to support the entire web application development process, from publishing and coding to testing and operations management. It is an AI agent-based user interface and user experience (UI/UX) development platform.

The company highlighted spec-driven development (SDD), which generates and validates outputs based on development specifications, as a core part of its AI development strategy. The approach is aimed at building an enterprise development environment that goes beyond code generation to manage development quality and reliability.

Inswave has supplied web standards technologies to about 900 customers across more than 4,000 projects in South Korea. The company said this experience allows it to apply AI-based development technologies in areas such as finance and the public sector, where stability and security are critical.

Inswave plans to hold follow-up discussions with local IT companies and potential customers through its Japanese subsidiary, Inswave Japan.

“We will use Inswave Japan as a foothold to expand engagement with local partners and create tangible business cooperation opportunities in the Japanese market,” said Eo Se-ryong, chief executive officer of Inswave.

Artificial intelligence (AI) services and solutions provider Bespin Global has won a project to upgrade "ChaTi," a conversational information search service operated by the Korea Agro-Fisheries & Food Trade Corp. (aT). The project will revamp the conversational search service within KATI, aT's specialized platform for agricultural and food export information, using the latest technologies.

Bespin Global said it will use its HelpNow AI platform to build a next-generation intelligent assistant based on agentic AI that analyzes user requirements and executes tasks. The goal is to help exporters access export information more conveniently and make faster, more informed decisions.

The project comes as the expansion of K-food exports increases the amount of information exporters need to regularly check, including country-specific quarantine requirements and labeling regulations. K-food exports reached a record $13.62 billion last year, of which agricultural and food products accounted for $10.41 billion. The government has set an export target of $16 billion for this year.

Bespin Global plans to improve ChaTi's answer accuracy by upgrading its latest large language model (LLM) and retrieval-augmented generation (RAG) architecture, expand training on internal data including export statistics databases and non-tariff barriers, and introduce an optical character recognition (OCR)-based system for extracting unstructured data from PDFs, tables and images.

In particular, the company will apply multi-agent orchestration technology that enables specialized agents in areas including market analysis, regulatory assessment and data analysis to collaborate on complex queries.

An incident in which OpenAI artificial intelligence (AI) agents circumvented developer-imposed controls and effectively took over a German wiki has become the subject of regulatory scrutiny by the European Union.

The European Commission (EC) has received an incident report from OpenAI and is reviewing the matter. Key questions include when OpenAI became aware of the agents’ abnormal behavior and whether it should have classified and promptly reported the event as a “serious incident” under the EU AI Act.

If OpenAI is found to have known that the incident was reportable but unjustifiably delayed reporting it, the company could face a fine of up to 3% of its global annual revenue.

The case has escalated into an EU regulatory issue because of the significant gap between when OpenAI may have learned of the incident and when it disclosed it publicly. According to Reuters, anonymous OpenAI employees said company executives knew about the incident weeks before Reuters first reported it.

Researchers at AI safety nonprofit Nightingale, which first uncovered the incident, also raised the possibility that OpenAI became aware of the situation around June 21. According to the researchers, AI agent activity on DSEwiki dropped sharply from June 21, shortly after a visit that appeared to be linked to OpenAI.

Reuters first reported the OpenAI agents’ activity on DSEwiki on Sept. 4.

From May through July this year, OpenAI’s AI agents made more than 15,000 edits to DSEwiki, a German wiki for programmers, using names such as “OpenAIResearcher” and “OAIResearchMar26.” Posts included answers to tasks being performed by the agents, methods for bypassing OpenAI safeguards and ways to evade monitoring.

DSEwiki is a privately operated site created about 25 years ago that had been largely abandoned for years. When its administrator belatedly discovered the unusual activity and began deleting posts, some agents identified the alphabetical deletion pattern and responded by creating backup pages beginning with the letter “Z.”

OpenAI maintains that the event was a “misalignment,” rather than a security incident. The company said security incidents go through formal reporting procedures, while misalignment is disclosed through system cards or research papers, and that there is no industry standard for separately reporting this type of AI behavior.

The first issue for the EC, therefore, is whether the DSEwiki incident qualifies as a “serious incident” under the EU AI Act.

Article 3(49) of the EU AI Act defines a serious incident as an incident or malfunction of an AI system that directly or indirectly results in death or serious harm to health, a serious and irreversible disruption to the management or operation of critical infrastructure, infringement of obligations under EU law intended to protect fundamental rights, or serious harm to property or the environment.

The second issue is the legal status of the AI model used by the agents. The EU AI Act exempts some general-purpose AI models that are in research, development or prototyping before being placed on the market from certain regulatory requirements.

The third issue is when OpenAI became aware of the incident and when it reported it. The EU AI Act requires providers of general-purpose AI models with systemic risk to report serious incidents “without undue delay.” It does not specify a fixed reporting deadline.

Based on currently available information, it is unclear exactly when OpenAI’s internal monitoring systems detected the incident.

Key EU obligations for providers of general-purpose AI models took effect on Aug. 2, 2025, while the EC’s related enforcement powers became fully applicable on Aug. 2 this year. General-purpose AI model providers are required to prepare technical documentation, publish summaries of training data and take measures to comply with copyright law.

The maximum fine for violations or the provision of incomplete information is 3% of global annual revenue or €15 million, whichever is higher. Prohibited AI practices can carry fines of up to 7% of global annual revenue or €35 million.

Political pressure could also increase in the US. Following the Hugging Face hacking incident in July, two US lawmakers asked OpenAI whether there had been additional similar cases, but the company reportedly declined to answer.

Beyond the regulatory questions, the DSEwiki incident also exposed concerns over the controllability of AI agents.

The agents were performing web retrieval tasks at the time. They were permitted to read information from the internet, but operated in an environment where the ability to write data was blocked. The agents, however, discovered that DSEwiki’s outdated software allowed edits through ordinary read requests, enabling them to circumvent the restriction on writing data.

The agents had not been instructed to communicate or collaborate with one another, but they used DSEwiki to exchange messages. When the administrator deleted posts, they analyzed the deletion pattern and created backup pages, adapting their behavior in response to human intervention.

Taken together, the AI agents circumvented restrictions explicitly imposed by developers, used the channel they gained to share information with other agents and changed their behavioral strategies in response to actions by a human administrator. In effect, they moved beyond human control.

Kakao Games will launch “Dokkaebi’s World,” a new massively multiplayer online role-playing game (MMORPG) developed by Supercat, on Oct. 8.

Kakao Games said Sept. 9 that it held an online showcase on Sept. 8, unveiling the game’s key content, service policies and launch schedule.

Dokkaebi’s World is a 2.5D MMORPG based on the intellectual property (IP) of Naver web novel and webtoon “Myeolgwi Sudojeon.” It combines 2D pixel-art characters with 3D backgrounds and features a world inspired by traditional Korean themes. Webtoon production company Redice Studio also participated in developing the IP.

The game’s main combat system, “Dosul,” consists of 12 styles. Each Dosul is linked to a unique mechanic called “Sinsik Gaebang,” allowing players to use different approaches depending on the combat situation.

The progression system emphasizes cooperation rather than competition among players, while incorporating optional player-versus-player (PvP) combat. Kakao Games also unveiled operational features including data management using an AI system called “Myorong” and “Dokkaebi General Assembly,” a program designed for regular communication with players.

Ahead of the official launch, Kakao Games will hold a live demonstration on Sept. 18. The developers will showcase actual gameplay and explain key content and systems in real time.

LG CNS is applying artificial intelligence (AI) agents to the Chaum health screening center. The company will build a system that analyzes customers' past screening records, recommends tests, drafts screening reports and supports follow-up care.

LG CNS said Sept. 9 that it recently held a kickoff meeting with Chaum and CHA Healthcare at the Chaum health screening center in Seoul's Gangnam district and began a proof of concept (PoC) for AI agents.

LG CNS will develop AI agents specialized for health screening operations, while Chaum will evaluate their efficiency and potential use in actual screening workflows.

Chaum currently creates personalized screening programs based on customers' past screening records, consultation history and physicians' recommendations. Nurses manually retrieve relevant information from multiple systems, including customer relationship management (CRM) and electronic medical records (EMR), while also preparing consultation records, managing follow-up care and scheduling.

LG CNS plans to have AI agents integrate and analyze screening records, test results, consultation records and follow-up schedules distributed across CRM and EMR systems. Based on previous screening results and medical staff recommendations, the AI will suggest necessary tests and explain the reasons for them. It will also help prepare materials explaining to customers why the tests are needed.

After screening, the AI will combine test results and medical assessments to draft screening reports. LG CNS also plans to use AI to prepare pre-screening precaution checklists, manage retesting and follow-up schedules at three-, six- and nine-month intervals, connect customers with abnormal findings to outpatient care and provide information on follow-up programs.

The companies are also considering introducing AI-powered interpretation and translation for foreign customers. Foreigners account for about 30% of customers at the Chaum health screening center, and AI could be used to provide screening results and assist during consultations.

LG CNS plans to use the PoC to establish an AI agent model specialized for health screening consultations and then expand its deployment to major CHA Medical Center health screening centers. LG CNS made a strategic investment in CHA Bio Group in January.

Naver has proposed expanding cooperation with French companies and institutions across artificial intelligence (AI) and other businesses.

Naver said Sept. 9 that CEO Choi Soo-yeon and Naver Cloud CEO Kim Yu-won met with French companies during South Korean President Lee Jae Myung's visit to France to discuss cooperation in AI infrastructure and content.

At a South Korea-France business roundtable held Sept. 8, Choi highlighted potential cooperation in three areas: global AI infrastructure, industry-specific AI with data sovereignty and security, and cultural content.

Choi noted that Naver recently announced plans with Nvidia and Brookfield to build more than 1 gigawatt of global AI factories. "We want to combine France's stable energy base with Naver's experience in the AI ecosystem to jointly build AI factories that can be used by European companies and startups," she said.

Choi also proposed supporting French companies in building industry-specific AI, citing Naver's financial AI "BOKI," developed with the Bank of Korea, and "Hy-NuRI," an AI system specialized for nuclear power plants and isolated from external networks, developed with Korea Hydro & Nuclear Power.

She also highlighted Naver Webtoon's business in France, where it has secured 2 million readers. "Based on our global webtoon platform and technological capabilities such as Toon Radar, our AI-powered copyright protection technology, we want to cooperate so that works created jointly by creators from both countries can reach readers around the world," Choi said.

Naver also signed a strategic memorandum of understanding (MOU) with France's National Audiovisual Institute (INA) to combine AI technology with audiovisual cultural heritage archives. The two sides plan to explore ways to apply Naver's AI technology to INA's extensive audiovisual archives to efficiently analyze and organize content and make it easier for users to find and use the materials they need.

Enterprise software support provider Rimini Street said Sept. 9 that healthcare company Ilhwa has adopted Rimini Support, its maintenance service for SAP S/4HANA systems.

Ilhwa plans to reduce maintenance costs through Rimini Street’s SAP S/4HANA technical support and use the IT budget savings to develop reports within its SAP system, fund business improvement projects and deploy new hardware.

Ilhwa, which operates mainly in the food and beverage and pharmaceutical businesses, has integrated key companywide processes including finance, sales, production and inventory around SAP S/4HANA. The system has also been customized to reflect the requirements of its various business units, making stable enterprise resource planning (ERP) operations and efficient maintenance important.

Rimini Street said Ilhwa needed a way to reduce its high annual maintenance costs while retaining its existing SAP S/4HANA environment and use its limited IT budget more strategically. Ilhwa also found that its previous maintenance arrangement fell short in resolving and responding to issues relative to its high cost.

Rimini Street provides Rimini Support L4 services for Ilhwa’s SAP S/4HANA environment, with dedicated engineers providing customized technical support. The company said Ilhwa has reduced its previous SAP maintenance costs by more than 50% since switching to its service.

Rimini Street also provides analysis reports on Ilhwa’s SAP system to improve operational efficiency. Ilhwa has reinvested part of the savings in small-scale business improvement projects, giving it greater flexibility to respond quickly to the needs of business units.

Ilhwa also cited having a single point of contact for maintenance support as a benefit. The arrangement allows it to resolve complex technical issues more quickly through dedicated Rimini Street engineers without having to go through multiple support channels.

Meta has announced its answer to OpenClaw. The company on Sept. 8 launched Muse, an agentic artificial intelligence personal assistant, in the U.S.

Muse is initially available through a dedicated app and WhatsApp. It offers a free basic tier as well as paid subscriptions costing $20 or $100 a month depending on usage.

Reports that Meta was developing an agentic AI personal assistant similar to OpenClaw emerged in May. The Information reported that Meta was training a project code-named Hatch using Anthropic's Claude 4.6 model and planned to switch to its own Muse Spark model for the commercial launch.

According to Meta, Muse can automate tasks ranging from online shopping and buying movie tickets to booking appointments such as tennis lessons and even filling out permission forms for children's school field trips. It is designed to access apps across categories including email, calendars, payments, health, shopping and smart homes.

Muse can send emails, book travel, reduce utility bills, fill out forms, make plans, turn recipe videos into grocery lists, send party invitations and make payments. Payments are handled through Link, a service from U.S. fintech company Stripe.

Stripe Link is a digital wallet that simplifies online purchases using stored payment information and supports payments made by AI agents on behalf of users. Its agent wallet generates one-time card numbers so actual card details are not exposed to merchants. Muse also receives Link's purchase protection.

How OpenClaw and Meta Muse Differ

Muse offers many features similar to OpenClaw, which served as a model for its development. One of OpenClaw's key strengths is its ability to carry out tasks when users simply send it a message through chat apps such as WhatsApp or Telegram. Muse implements a similar approach through its dedicated app and WhatsApp.

Another similarity is its always-on capability. Muse runs independently on a dedicated virtual machine (VM) in the cloud regardless of whether the user's computer is turned on. This provides the same always-on capability that allows OpenClaw to independently perform monitoring and scheduled tasks even when a user's computer is off.

Muse also resembles OpenClaw in the breadth of its integrations and autonomous actions. It can send emails, book travel, make payments and even negotiate. Sensitive actions such as sending emails or making purchases require user approval, but other tasks can be completed autonomously without requiring human confirmation at every step. That mirrors OpenClaw's ability to complete an entire workflow from a single message.

OpenClaw, however, is open source, allowing users to install it on their own servers and connect models of their choice. Muse is a commercial cloud service that runs only on Meta's Muse Spark model. The biggest difference is security.

OpenClaw requires users to manage security themselves and has faced numerous concerns over vulnerabilities. Muse, by contrast, incorporates privacy and security safeguards from the design stage, including isolated secure VMs and separate connection approvals for individual apps. In that respect, it is similar to Nvidia's NemoClaw, which combines OpenClaw with a security framework.

NemoClaw for Enterprises and Developers, Muse for Consumers

Nvidia's NemoClaw is an open-source security and deployment stack designed to run OpenClaw securely. OpenClaw serves as the brain, while NemoClaw acts as protective headgear around it.

NemoClaw uses a sandbox runtime called OpenShell to block and monitor access to file systems, networks and processes, preventing OpenClaw agents from communicating externally without authorization. It also incorporates security and privacy policies, network access controls, credential management, model routing, agent lifecycle management and observability.

Nvidia's own language models, including Nemotron, can also run directly on a user's PC in a local environment. This allows inference and execution without sending data to the cloud. The configuration can be version-controlled through YAML policy files, allowing multiple developers to reproduce the same security settings.

Muse and NemoClaw share the concept of long-running agents that autonomously perform tasks. Both make isolated execution environments that prevent user data from being exposed externally a core design principle — cloud VMs for Muse and local sandboxes for NemoClaw.

Both approaches give AI agents permission to operate web browsers and apps while establishing guardrails to prevent unauthorized access to the broader system or sensitive information.

OpenClaw and NemoClaw target developers and users with some AI and IT expertise, while Meta's Muse can be seen as extending the concept to general consumers.

Always-On Operation and Long-Term Memory Key to Autonomous AI Assistants

Agentic personal AI assistants require computer-use technology, always-on operation and long-term memory, as well as broad web and app support.

Computer use is a fundamental technology for agentic personal AI assistants. It allows AI to perceive a screen and operate a computer like a person, moving the cursor, clicking buttons, entering text and navigating websites and apps. It effectively provides the AI with eyes and hands, with the advantage of not requiring separate integrations for every app.

While computer use is essential, always-on operation is even more important. Users need to be able to issue instructions and interact with the assistant at any time, while continuously pursuing long-term goals that may span several days requires a separate orchestration layer.

The assistant must also remember and preserve users' work and include dedicated connectors, or application programming interfaces (APIs), for functions such as payments and calendars to operate reliably in real-world tasks.

An agentic personal AI assistant therefore needs more than computer-use capabilities. It requires an always-available entry point through messaging services, long-running background execution that continues when the user is away, persistent memory of previous conversations and preferences, and an ecosystem of connectors spanning services such as email, payments and calendars.

By those criteria, services that can currently be considered agentic personal AI assistants are largely limited to OpenClaw, Meta Muse and Perplexity's Computer.

OpenAI and Anthropic, two leading AI research companies, do not yet offer fully autonomous personal AI assistants. Both companies are primarily focused on enterprises and developers.

Google continues to add agentic AI assistant features to its Gemini app, but its strengths remain largely within Google's ecosystem, including Gmail, Calendar and Docs. Microsoft Copilot is similar, with a focus on work environments built around Microsoft 365 applications such as Outlook, Word and Excel.

A South Korean court granted a divorce petition filed by the spouse of Smilegate founder and Chief Visionary Officer (CVO) Kwon Hyuk-bin. The court recognized the spouse’s contribution to the couple’s assets at 35%, resulting in a property division worth 2.55 trillion won, the largest in a South Korean divorce case.

The Seoul Family Court’s Family Division 3 on Sept. 9 granted the divorce of Kwon and his wife, identified by her surname Lee, and divided their assets 65% to Kwon and 35% to Lee.

Considering that most of Kwon’s assets are unlisted shares, the court ordered 35% of the shares to be transferred in kind and an additional 65 billion won to be paid in cash to cover the shortfall. The court assessed Kwon’s assets at about 7.34 trillion won, including about 7.10 trillion won worth of Smilegate shares. Lee was also designated as the legal custodian and caregiver of their child. Kwon must pay 20 million won in monthly child support at the end of each month until the minor child reaches adulthood.

A key issue in the lawsuit was how much of Kwon’s Smilegate holdings should be subject to property division. Kwon owns 100% of Smilegate. Because shares in the privately held company account for most of his assets, the size of the settlement depended heavily on how the court valued the stake.

Lee had initially sought half of Kwon’s Smilegate stake. She married Kwon in 2001, a year before Smilegate was founded. Lee argued that she held shares and participated in management during the company’s early years and later contributed to the accumulation of assets through household responsibilities and raising their children. Kwon maintained that Lee was not a co-founder and that her contribution to the company’s growth was relatively limited.

By the size of the property division, the ruling is the largest in a divorce involving a South Korean business executive. In an earlier retrial following a Supreme Court remand in the divorce case between SK Group Chairman Chey Tae-won and Art Center Nabi Director Roh Soh-yeong, the assets were divided one-third to Roh and two-thirds to Chey, with Chey ordered to pay 944 billion won.

Kwon’s lawyer said the legal team would review the ruling and decide on further proceedings in consultation with Kwon.

Intelligent power management company Eaton unveiled its next-generation data center power infrastructure strategy and key technologies at Eaton Data Center Tech Day 2026, held at the Atlas Hall of The Westin Seoul Parnas on Sept. 10.

Eaton's vice president of engineering for its Electrical Sector in Asia Pacific introduced a "power-to-chip" ecosystem and infrastructure design strategy for next-generation data centers. The company presented solutions including its 9395XR uninterruptible power supply (UPS), distributed power quality (DPQ) portfolio, AI data center cooling systems, prefabricated white space, grid interconnection and power distribution solutions, and electrical power monitoring systems (EPMS) to support the design, construction and operation of AI data centers.

A key Eaton solution is its medium-voltage solid-state transformer (MVSST). Based on advanced power electronics technology, the MVSST simplifies the conversion of high-voltage alternating current (AC) to direct current (DC), supporting 800VDC AI data center architectures. It is designed to reduce energy conversion losses and system complexity while improving reliability.

The MVSST provides system efficiency of up to 98.5% and power density of up to 277.7 kilowatts per square meter. Integrating medium- and low-voltage cabinets with converters can reduce installation space by as much as 50%, while simplified system configurations and prefabrication can cut overall project lead times by as much as 50%.

The MVSST also uses bidirectional power conversion to serve as an intelligent energy router, flexibly connecting the power grid with renewable energy sources such as solar and wind, energy storage systems (ESS) and DC microgrids. This enables energy flows between power sources and loads to be managed efficiently according to demand while supporting stable operation in high-power, high-density environments.

As adoption of autonomous artificial intelligence, or agentic AI, and agentic coding accelerates, the industry is confronting a paradox: per-token costs are falling while total AI spending is surging.

Unlike one-off chatbot conversations, agentic AI autonomously repeats reasoning, planning, tool calling, verification and error-correction loops. Estimates suggest agentic AI could consume five to 30 times as many tokens as a single query, and as much as 1,000 times more for complex tasks.

Recently released AI models are increasingly promoting token efficiency as a competitive advantage. Beyond lowering the cost per token, developers are emphasizing technologies that reduce the number of tokens used for reasoning itself. Prompt caching, batch application programming interfaces (APIs) and dynamic tool loading have been used to cut token-processing costs and eliminate unnecessary context.

More advanced approaches to token optimization are now emerging. The most prominent is latent-space reasoning, along with more aggressive approaches such as Soft Latent Thinking and Selective Latent Thinking, and the Full-Bandwidth Transformer, which seeks greater token efficiency while retaining the existing architecture.

Token Compression Does Not Eliminate Inefficiency

Competition among large language models (LLMs) is shifting from more parameters toward more efficient reasoning. The ability to solve the same problem with fewer tokens is becoming a new competitive metric.

Models with stronger reasoning capabilities tend to generate more reasoning tokens before producing an answer. Longer reasoning can improve accuracy, but generating more tokens increases decoding operations and graphics processing unit (GPU) computation. It also expands the KV cache used to store information from previous tokens. Performing the same reasoning with fewer tokens therefore directly affects AI model costs and latency.

Heima, or Efficient Reasoning with Hidden Thinking, presented at the International Conference on Machine Learning (ICML) in April, approaches the problem through compression. It reduces the number of tokens used for reasoning by compressing long chains of thought (CoT) or eliminating unnecessary reasoning steps.

Heima sharply reduces the length of "thinking tokens" used internally by a model to analyze and reason before generating its final answer. It compresses lengthy, complex reasoning into a handful of core tokens, which can be converted back into human-readable text by a dedicated Adaptive Interpreter when needed. The approach preserves how the model thinks while reducing how extensively that thinking is expressed in language.

Token compression can immediately reduce API costs without changing the underlying model architecture. But even with highly effective compression, reasoning is still expressed as natural-language tokens.

No matter how complex the model's internal calculations are, information passed to the next step must ultimately travel through the narrow channel of a single selected token. This is an inherent limitation of autoregressive decoding.

Autoregressive decoding is how transformer-based LLMs generate text sequentially, one token at a time, using previously generated tokens as inputs for the next step. Because the model continuously references earlier context, it can maintain coherent flow, tone and grammar even across long passages.

The approach, however, has clear limitations. During reasoning, each new token depends on previously generated tokens, making token generation inherently sequential and difficult to parallelize.

An incorrect token generated midway can also cause errors to accumulate through subsequent computations, distorting the context. Each new token must pass through the model's large neural network layers, adding computational costs and memory bottlenecks.

Latent Reasoning Moves Beyond Autoregressive Decoding

The limitation stems from a gap between how models think and how they transmit information.

Transformers process input tokens through multiple layers and transform them into high-dimensional hidden states containing thousands of dimensions. A hidden state represents information exchanged internally between model layers and is not exposed outside the model. It can be compared with an intermediate note the AI keeps internally before expressing an answer.

These hidden states contain not only context and meaning but also rich latent information such as incomplete reasoning, partial answers and future plans that have yet to be fully expressed as words. The problem is that this high-dimensional information is not passed directly into the next computational step.

Instead, the model uses the hidden state to calculate a probability distribution for the next token, selects one token and uses it as the input for the next step. The hidden state itself is not directly carried forward.

As a result, even sophisticated internal computations must pass through an extremely narrow bottleneck of a single word. Rich high-dimensional information represented as continuous vectors is compressed into a language token and then converted back into a vector for the next calculation, creating repeated information loss and computational inefficiency.

The problem goes further. Every time the model selects the next token, it projects the hidden state across the entire vocabulary to calculate probabilities for a large number of candidate tokens. This process, known as LM Head Projection, imposes a substantial computational burden.

Adding a single word therefore requires the model to repeatedly perform a large-scale probability calculation across tens or hundreds of thousands of vocabulary candidates. Reducing the number of tokens does not eliminate this fundamental computational structure.

Latent reasoning, which has recently gained attention as a token-optimization technology, seeks to fundamentally change part of this cost structure.

Does Intermediate Reasoning Need to Be Expressed in Natural Language?

Latent reasoning performs intermediate reasoning directly within the model's continuous internal state, rather than expressing it in natural language. In human terms, instead of verbalizing every intermediate thought before reaching a conclusion, the model quietly works through the problem internally.

Traditional reasoning incurs costs by converting internal thoughts into language tokens and then decoding those tokens back into representations for further computation. Latent reasoning instead completes intermediate calculations within the model's hidden numerical space, or latent state, without generating each thought as human-readable natural language.

One example pushing this concept further is the BDH-CQ model unveiled last month by Pathway Research, part of Pathway. Pathway develops post-transformer architectures and efficient AI reasoning technologies aimed at overcoming limitations of transformer models.

BDH-CQ has only 150 million parameters but attracted industry attention for its performance relative to its size on ARC-AGI-1, a well-known benchmark for difficult AI reasoning tasks.

ARC-AGI-1 evaluates a model's ability to infer and solve previously unseen spatial grid-pattern rules. It tests whether a model can independently construct geometric and spatial logic, similar to a child figuring out the rules of a new toy.

While conventional AI benchmarks often measure memorized knowledge or language understanding, ARC-AGI tests how quickly a model can identify and apply unfamiliar rules without prior knowledge. Models get only two attempts, or Pass@2, and even a single incorrect pixel results in a wrong answer.

BDH-CQ scored 29.5% under Pass@2. Frontier models perform far better under the same conditions, with OpenAI's GPT-6 Astra at 98.5%, Anthropic's Claude Opus 5 and Fable 5.1 at 97.5%, and Google Gemini 3.7 Flash at 95.5%.

BDH-CQ's 29.5% appears low by comparison, but its parameter count is far smaller. While parameter counts for frontier models are undisclosed, they are estimated at at least 1 trillion and are believed to be around 2 trillion to 5 trillion. BDH-CQ achieved nearly 30% with roughly one ten-thousandth as many parameters. Its compute cost per task was also just $0.0007.

150-Million-Parameter BDH-CQ Nears ARC-AGI-1 Pareto Frontier

The potential impact could be significant if Pathway Research can further improve its latent reasoning-based BDH-CQ model.

BDH-CQ does not output intermediate reasoning in human language tokens. Instead, it repeatedly updates its recurrent memory and performs computations entirely within its internal latent-state space.

For AI developers, reducing or eliminating the process of converting internal thoughts into language — projecting hidden states across the vocabulary through the LM head, selecting the next token and embedding that token again for the next reasoning step — could enable comparable performance with less computation. That could ultimately deliver high performance efficiency even in smaller models.

COCONUT demonstrated the possibility that AI can reason without expressing thoughts in language, while BDH-CQ extended the concept toward smaller models and lower inference costs.

Both approaches reduce the role of language tokens in AI reasoning and directly use the model's continuous internal states. Researchers are now going a step further by exploring whether language generation itself can be removed while an AI model is thinking. One such approach is Soft Latent Thinking.

A paper titled "A Model with No Head and Many Thoughts" appeared on arXiv on Aug. 31. Its authors are researchers from Russian internet and technology company Yandex and Russian financial technology company T-Tech.

The Soft Latent Thinking method described in the paper removes the LM head, the language model's output mechanism, during reasoning and replaces it with a lightweight projector. The projector transforms hidden states into continuous representations in embedding space and passes them directly into the next reasoning step.

Soft Latent Thinking Can Be Applied to Existing LLMs

Conventional language models repeatedly convert hidden states representing their internal thoughts into actual language tokens during intermediate reasoning. The model calculates probabilities across a large vocabulary containing tens of thousands of token candidates, selects the next token and uses it as the input for the next reasoning step.

The mechanism that converts hidden states into vocabulary-space probabilities is called the Language Model Head, or LM head.

Soft Latent Thinking eliminates this intermediate language-generation process. A much lighter projector directly connects the current latent state to the next one. The model continuously updates its vector state without generating language tokens while reasoning, converting it back into language only when producing the final answer.

BDH-CQ builds recurrent latent reasoning into a small 150-million-parameter model, while Soft Latent Thinking differs in that it can also be applied to existing LLMs.

Selective Latent Thinking

Soft Latent Thinking reduces computation by eliminating language-token generation during reasoning and processing the entire thought process in latent space. But that raises another question: Is compressing every reasoning step into a latent state always optimal?

Compressing stages that require detailed intermediate reasoning, such as complex mathematical calculations or precise logical deduction, could instead reduce accuracy.

Research is therefore expanding toward approaches that distinguish between reasoning segments that can be compressed and those that should remain in language, rather than converting the entire reasoning process into latent representations. Selective Latent Thinking is a representative approach.

Selective Latent Thinking predicts upcoming short reasoning segments and uses the model's confidence level to determine whether they should be compressed into latent representations. Segments that can be safely compressed are skipped through latent states, while uncertain sections or those requiring precise reasoning retain explicit CoT. The idea is that not every thought needs to be expressed in language, nor does every thought need to be moved into latent space.

A Selective Latent Thinking paper published in May tested the approach using Llama-3.2-1B and Qwen3-4B across four mathematical reasoning benchmarks.

The researchers reduced reasoning-chain length by 58.4% compared with explicit CoT while limiting the decline in accuracy to 2.8%. In other words, a reasoning process equivalent to 100 tokens could be simplified to about 42 tokens while largely maintaining performance.

The researchers also found that Selective Latent Thinking improved the balance between accuracy and efficiency compared with conventional latent reasoning. At similar levels of reasoning compression, it achieved 22.7% higher accuracy than existing latent reasoning methods.

Microsoft Studies Full-Bandwidth Transformer

Microsoft AI Frontier researchers' Full-Bandwidth Transformer retains generated tokens as in conventional models but also passes the previous step's hidden state into the next step, increasing the amount of information available during the same reasoning process.

The research, released last month, retains the standard LLM transformer architecture, KV cache and existing language-modeling objective while widening the information channel between steps. It combines the previous step's top-level hidden state with the embedding of the newly generated token using a Gated Linear Unit (GLU) and passes both into the next step.

According to the research, a 1-billion-parameter Full-Bandwidth Transformer trained on 400 billion tokens achieved performance comparable to a conventional transformer trained on about 600 billion tokens, or roughly 1.5 times as much data. This suggests models of the same size can achieve similar performance with less training data.

Differences also emerged during inference. The reasoning trace required to reach the same level of accuracy was shorter.

In other words, reusing information the model has already calculated rather than discarding it can deliver similar performance with less data during training and shorten the reasoning trace required to achieve the same accuracy during inference.

From Parameter Scaling to Reasoning Efficiency

Taken together, recent AI reasoning-efficiency technologies including Heima, COCONUT and BDH-CQ are moving toward changing the way models think. Heima compresses long reasoning processes into smaller thinking tokens, while COCONUT and BDH-CQ perform reasoning in latent space rather than human-readable language.

Soft Latent Thinking minimizes language generation during reasoning, while Selective Latent Thinking retains language only where needed and compresses the rest. Microsoft's Full-Bandwidth Transformer, meanwhile, preserves the token-based architecture while carrying hidden states into the next step so the model can use more information during the same reasoning process.

The approaches differ, but their direction is the same: improving AI performance increasingly involves changing how models internally process information and reason.

The AI market is likely to evolve toward competition not only in raw performance but also in how much of that performance can be achieved with fewer tokens and less computation.

AI and data infrastructure solutions provider HS Hyosung Information Systems is targeting enterprise AI operations with high-efficiency inference infrastructure combining its low-power Arm-based GreenCore server with neural processing units (NPUs).

HS Hyosung Information Systems said Sept. 9 it is working with major domestic and global NPU companies including Qualcomm, FuriosaAI, DEEPX and Mobilint to provide infrastructure suited to customer environments. It also offers its Arm-based GreenCore server as an option to maximize NPU performance. GreenCore was jointly developed by Korean Arm server manufacturer ACCESSLAB and HS Hyosung Information Systems.

The company describes GreenCore as an Arm architecture-based central processing unit (CPU) server featuring low power consumption, low heat generation and efficient many-core parallel processing. It is designed for data centers with limited power, cooling and rack space, as well as companies and institutions preparing on-premises AI services due to security and data sovereignty requirements.

“NPUs are gaining attention as an alternative to GPUs, but supported AI models and target workloads differ depending on the NPU manufacturer and product,” the company said. “It is therefore important to select products suited not only to the characteristics of the service but also to deployment environments such as data centers and edge infrastructure.”

HS Hyosung Information Systems analyzes customers’ AI models, service types, model sizes, concurrent user counts, target response times and data center power and space conditions to recommend suitable NPU candidates.

Once an NPU is selected, the company installs it in a GreenCore server and conducts integrated validation covering hardware and software compatibility, AI model execution and operational stability before deploying the system in the customer’s environment.

When needed, the company also works with AI solution providers to conduct proof-of-concept (PoC) testing that reflects actual service environments. This reduces the burden on customers of directly comparing and validating multiple NPU products and vendor-specific development tools. HS Hyosung Information Systems supports the entire AI inference infrastructure deployment process, from workload analysis and NPU selection to GreenCore server configuration, integrated validation, AI solution integration, PoC, deployment and operations.

The company plans to continue expanding its HS Hyosung AI Platform portfolio by integrating Arm servers, GPUs, NPUs, high-performance storage, data lakes, artificial intelligence for IT operations (AIOps) and its AI solution partner ecosystem.

Alphabet's Google will invest €13 billion in artificial intelligence (AI) data centers and related digital infrastructure in Finland, its largest single investment in Europe. The investment is supported by the country's cold climate, abundant water resources and electricity supply from nuclear power.

Google said it will work with Finnish municipalities, telecommunications companies, grid operators and other local organizations and businesses. The company has operated a data center in Hamina since converting a former paper mill into the facility in 2009.

The investment will expand the existing data center and add new facilities in Kajaani, Muhos and Vaala in northern Finland, where access to the power grid is strong.

Nordic countries including Finland and Sweden have emerged as attractive locations for new AI data centers because of their cold climates, abundant water resources from seas and lakes, and clean nuclear and renewable energy such as hydropower and wind. These conditions can reduce the energy consumed by cooling systems, which account for a significant share of data center electricity use.

According to a U.S. Department of Energy (DOE) energy-efficiency study, cooling systems account for about 40% of energy consumption at traditional data centers. Gartner recently estimated that cooling and other infrastructure account for about 35% of overall data center electricity consumption.

A Gartner report released in June projected that electricity consumption by cooling and other infrastructure will rise sharply from 159 TWh in 2025 to 195 TWh in 2026 as the number of dedicated AI servers increases.

The Nordic region's low outdoor temperatures allow data centers to extend the use of free cooling, reducing reliance on mechanical cooling. Lower utilization of air conditioners, chillers and other mechanical cooling equipment can significantly reduce cooling energy costs.

Abundant water resources also enable direct water cooling, which draws cold seawater or lake water through pipes to absorb server heat before discharging it. The method can rapidly cool large-scale servers without compressors or cooling towers. Google already uses a seawater-based cooling system at its Hamina data center, drawing cold water from the Baltic Sea to cool servers.

Abundant electricity is another reason hyperscalers are turning to the Nordic region. Hydropower in Norway and wind and nuclear power in Finland and Sweden can help hyperscalers reduce carbon emissions. As part of its Finland investment, Google has entered into a carbon-free energy partnership with Finnish state-owned energy company Fortum Oyj.

Google signed a 22-year agreement covering up to 50% of the capacity of Fortum's Loviisa nuclear power plant. The agreement will start at a smaller scale in 2028 and reach 50% of the plant's capacity from 2030 through 2049. It is Google's first nuclear power procurement agreement outside the U.S. The company also signed power purchase agreements totaling 629 MW with onshore wind developers including Valorem and Suomen Hyötytuuli.

Google will also work with Finnish transmission system operator Fingrid Oyj to ensure reliable grid connections and support grid stability. A 94-MW battery energy storage system (BESS) is planned near its Kajaani site to help balance the grid. The company will also work with Nokia and telecommunications service provider Elisa on high-speed network connectivity and other communications infrastructure.

Microsoft is similarly building an AI data center near Helsinki after investments in Sweden and Denmark. It is using the Nordic region's cold climate to improve cooling efficiency while planning to feed waste heat generated by AI computing into district heating networks for local residents.

Amazon Web Services (AWS) is modernizing and expanding its cloud infrastructure in Sweden, including Eskilstuna, Katrineholm and Västerås, as well as in Finland, to accommodate the high power density required by AI infrastructure. AWS has long-term wind power purchase agreements (PPAs) in Finland and Sweden and combines them with cold-air cooling. Its data centers are reported to operate at a power usage effectiveness (PUE) of 1.1 or lower.

Meanwhile, Argentina's Patagonia region, which has similar climatic conditions, is also emerging as a potential location for AI data centers.

OpenAI has signed a letter of intent (LOI) with local energy company Sur Energy to consider a green data center project worth up to $25 billion with capacity of as much as 500 MW. A lack of high-speed fiber-optic networks and transmission infrastructure, however, could require additional investment.

AI infrastructure company Snet Systems said on Sept. 10 it has completed a wireless network upgrade project at Kyung Hee University Medical Center, establishing infrastructure to support a smart hospital environment.

Kyung Hee University Medical Center launched the project to meet growing demand for digital healthcare services and smart medical environments. Snet Systems deployed a high-performance wireless network across major clinical and operational areas, providing a foundation for stable mobile electronic medical record (EMR) and medical Internet of Things (IoT) services.

The new network allows medical staff to use smart devices to immediately access and enter patient information at the point of care. It is expected to reduce errors and workloads associated with handwritten records and subsequent data entry, while real-time EMR connectivity supports faster and more accurate medical care.

The seamless wireless environment across key hospital areas is also expected to improve continuity of medical services and operational stability. Snet Systems said it carried out the project based on its experience building and maintaining the medical center’s existing Cisco-based wired network. Having a single provider operate both wired and wireless networks also enables faster responses to network disruptions.