Journalist

Kim Seong-hyeon
  • Anthropic and OpenAI Target Corporate and Government Markets Amid $700 Billion AI Infrastructure Investment
    Anthropic and OpenAI Target Corporate and Government Markets Amid $700 Billion AI Infrastructure Investment As Anthropic accelerates its push into corporate and public markets with major contracts with PwC and the Gates Foundation, OpenAI has opened its cutting-edge model to all levels of the U.S. government. The combined AI infrastructure investment from the four major tech companies this year is projected to reach a record $700 billion. Anthropic Secures Major Contracts with PwC and Gates Foundation According to the IT industry on May 17, PwC announced it will sequentially deploy its Claude code and Claude CoWork to teams in the U.S., expanding to hundreds of thousands of personnel worldwide. Initially, a certification program for 30,000 experts in the U.S. will be launched, along with the establishment of a joint excellence center. The collaboration will focus on developing software agents, executing AI-based deals, and transforming core business functions such as finance, supply chain, and HR. The reported improvement in delivery to clients could be as high as 70%. On May 15, Anthropic announced a four-year partnership worth $200 million (approximately 280 billion won) with the Gates Foundation. Anthropic will provide engineering personnel and API credits, while the foundation will handle grants and field expertise. The partnership aims to develop AI tools in health, education, and agriculture, including computational screening for vaccine candidates for neglected diseases like polio and HPV, and building K-12 education platforms in sub-Saharan Africa and India. This deal is four times larger than the $50 million contract signed between OpenAI and the Gates Foundation in January. OpenAI Opens Advanced Model to All Levels of U.S. Government OpenAI has opened its advanced AI model to all verification levels of the U.S. government. Last month, OpenAI launched its cyber defense-focused model, GPT-5.4-Cyber, and expanded its existing Trusted Access Cyber (TAC) program to federal, state, and local governments. This includes small operators of critical infrastructure such as hospitals, schools, and water facilities. Top-level certified users will have access to specialized versions with broader permissions for sensitive tasks like vulnerability research and analysis. OpenAI also hosted hands-on workshops for government representatives from the Department of Defense, Department of Homeland Security, and DARPA in Washington, D.C. At the same time, Microsoft, Google, and xAI signed a pre-launch model evaluation agreement with the AI Standards Innovation Center (CAISI) under the U.S. Department of Commerce. OpenAI and Anthropic also renegotiated their existing partnership in line with the AI action plan from the Trump administration. In contrast, Anthropic is limiting the distribution of its Mythos model to a consortium called Project Glasswing, which consists of about 40 approved agencies. Big Tech Companies to Invest Up to $700 Billion in AI Infrastructure This Year The combined capital expenditures (capex) of Amazon, Google, Meta, and Microsoft for 2026 are expected to reach between $630 billion and $700 billion (approximately 880 trillion to 980 trillion won). Amazon is projected to spend $200 billion, Alphabet between $175 billion and $185 billion, Meta between $115 billion and $135 billion, and Microsoft over $120 billion. This represents an increase of more than 60% compared to 2025. Approximately 75% of this spending will be directly allocated to AI infrastructure, including AI servers, GPUs, and data centers. Financial pressures are also mounting. Morgan Stanley forecasts that Amazon will record a cash flow deficit of about $17 billion this year. Barclays predicts that Meta's cash flow could decrease by as much as 90%. The four major tech companies have already issued bonds totaling $100 billion this year. Goldman Sachs estimates that the combined capex of hyperscalers will exceed $1.15 trillion over the three years from 2025 to 2027.* This article has been translated by AI. 2026-05-17 10:33:51
  • New AIDC Law Raises Concerns Over Renewable Energy Viability
    New AIDC Law Raises Concerns Over Renewable Energy Viability The Special Act on the Promotion of the Artificial Intelligence Data Center (AIDC) industry passed the National Assembly on May 7, drawing attention to Article 20, which allows direct trading of renewable energy. This provision enables AIDC operators to enter power purchase agreements (PPAs) directly with renewable energy producers, bypassing Korea Electric Power Corporation (KEPCO). The government aims to address the chronic issue of power procurement for AIDCs through renewable energy. However, industry reactions have been lukewarm. On May 14, multiple sources from domestic AIDC operators and construction firms expressed skepticism, stating that "direct trading of renewable energy is a provision that institutionalizes losses under the current structure." They criticized the legislation for not considering the critical role of electricity procurement costs in the operational profitability of AIDC businesses. AIDCs are extremely power-intensive facilities, consuming dozens of times more electricity than typical office buildings. Hyper-scale data centers require over 100MW of continuous power supply around the clock. Electricity costs account for 40-50% of total operating expenses, with AIDCs facing even higher burdens due to their power density. Operators calculate their break-even points differently based on size, but on average, small AIDCs under 40MW find profitability at around 150 won per kWh, while hyper-scale AIDCs over 100MW reportedly face losses if costs exceed 220 won per kWh. The problem is that the current market price for renewable energy in South Korea far exceeds these break-even points. According to industry sources, the current direct trading price for renewable energy, including solar power, is around 250 won per kWh, with offshore wind exceeding 300 won. In direct trading, prices are negotiated between suppliers and consumers, leading to increases of over 100 won per kWh compared to generation costs. The Korea Energy Economics Institute estimates the levelized cost of electricity (LCOE) for solar power in South Korea at 122 won per kWh and offshore wind at 238 won as of 2024. The actual supply prices are higher than the LCOE due to additional costs such as renewable energy certificates (RECs), grid connection fees, and backup power costs to address intermittency. South Korea's solar LCOE is reported to be 2.1 to 2.5 times higher than the global average, while onshore wind is about three times higher, and offshore wind is 1.3 to 2.4 times higher than in major countries. An official from an AIDC construction firm stated, "The moment we sign a renewable energy PPA, we confirm operational losses due to electricity costs. Even if global tech giants like Google and Microsoft enter Korea, they cannot achieve profitability under the current renewable energy supply prices." In the current landscape, AIDC operators see nuclear power as a realistic alternative. The current cost of nuclear power is around 60 to 80 won per kWh, significantly lower than renewable energy prices, at about one-third to one-fourth of the cost. However, Article 20 of the AIDC law only specifies direct trading of renewable energy. There is no provision for AIDC operators to purchase nuclear power directly, nor is there any law facilitating the supply of nuclear electricity. Nuclear power can still only be obtained through KEPCO, which applies its industrial electricity pricing structure. According to the Korea Energy Economics Institute, it is estimated that renewable energy will not reach the AIDC break-even point until after 2050. An industry source remarked, "The most suitable power source for AIDCs is stable nuclear energy supplied 24/7, yet the law only opens the door for direct trading of renewable energy. This provision is disconnected from reality for both global tech giants needing to meet RE100 requirements and domestic AIDC operators." 2026-05-14 16:23:02
  • AIDC Law Passed, But Power Line Issues Persist in Seoul Area
    AIDC Law Passed, But Power Line Issues Persist in Seoul Area The National Assembly passed the "Special Act on the Promotion of the AI Data Center Industry" (AIDC Law) on May 7, but industry experts are calling it a "toothless special law." As of May 14, the IT sector reported that from August 2024 to June 2025, approximately 290 applications for power usage at data centers were submitted nationwide, with 195 applications (67%) concentrated in the metropolitan area. While there is a high volume of power usage applications for AIDC in the metropolitan area, all hyper-scale AIDCs under construction nearby have yet to receive permits for power line and tower installations. The Ministry of Climate and Energy and local governments are responsible for granting these permits, and industry insiders express skepticism that the passage of the AIDC Law will resolve the transmission and distribution issues. The most significant barrier to AIDC development is the power infrastructure. Hyper-scale AIDCs, which require continuous operation for large-scale computations, need a stable power supply ranging from tens to hundreds of megawatts (MW). This necessitates the installation of power lines and substations. According to the IT industry, several AIDC construction companies and operators have indicated that the current stance of relevant authorities is clear: "Permits will not be granted unless the project is outside the metropolitan area." The consensus in the industry is that the only regions where hyper-scale power permits are effectively granted are non-metropolitan areas, including underprivileged regions. The metropolitan area is nearing saturation in terms of system capacity, and available power is also concentrated in non-metropolitan areas. The situation for AIDCs in the metropolitan area is even more dire. Article 19 of the AIDC Law grants exemptions from power system impact assessments under the "Special Act on the Activation of Distributed Energy" only for non-metropolitan areas. This exemption does not apply in the metropolitan area, where operators must still undergo power system impact assessments that take over 150 days. Even after completing the assessment, unless the authorities change their permitting stance, there will be no practical benefits. Article 24, which designates special zones for AI data centers, is also limited to non-metropolitan areas. Being designated as a special zone provides various benefits, including reductions in alternative forest resource creation fees, agricultural land preservation fees, and traffic impact fees, as well as priority guarantees from the credit guarantee fund. However, metropolitan area operators are excluded from these benefits, creating a direct conflict with the reality that 67% of nationwide data center power usage applications are concentrated in the metropolitan area. Although Article 4 of the law states that it takes precedence over other laws and Article 20 allows for direct transactions of renewable energy, this does not replace the permitting process for installing power lines and towers. The actual authority regarding power delivery remains within the frameworks of the Electricity Business Act and the Act on the Promotion of Power Development. Even if the AIDC Law legalizes direct transaction routes for electricity, without permits for laying down the lines, such contracts become meaningless. The direct transaction of renewable energy is also expected to lead to operating losses due to supply prices exceeding 200 won per kWh. Another significant flaw in the legislation is that most of its core provisions are delegated to presidential decrees. The minimum size criteria for supported AIDCs, the upper limit for power capacity exempt from non-metropolitan power system assessments, and the criteria for special zone designation are all left to the implementation decree. This means that the actual scope of benefits will depend on how the decree is formulated. The law will take effect nine months after its promulgation, meaning its actual impact will not be felt until after the end of this year. Concerns have also been raised that if the criteria for supported sizes are set too high during the decree consultation process, small and medium-sized AIDCs may again fall into a gap. Industry experts agree that resolving the power issues for AIDCs in the metropolitan area requires solutions outside the AIDC Law. Proposed measures include simplifying the installation procedures for AIDC power facilities through amendments to the Act on the Promotion of Power Development, mandating KEPCO's system connections, and allowing direct transactions for nuclear power. Analysts suggest that the current legal framework, which distributes jurisdiction over the AIDC Law, the Act on the Promotion of Power Development, and the Electricity Business Act among the Ministry of Science and ICT, the Ministry of Climate and Energy, and the Ministry of Trade, Industry and Energy, creates structural bottlenecks for investments in AIDCs in the metropolitan area. A representative from a company involved in AIDC construction near the metropolitan area pointed out, "While we can obtain building and transmission permits, we cannot lay down the lines, rendering the AIDC Law effectively meaningless." 2026-05-14 16:19:42
  • SK Telecom Hosts Forest Retreat for Long-Time Customers
    SK Telecom Hosts Forest Retreat for Long-Time Customers At 9 a.m. on May 10, a light green banner reading "Path to SK Telecom Forest Retreat" caught the eye along the road lined with trees at Hyangsusan in Yongin, Gyeonggi Province. This sign pointed the way to the Forest Camp at Everland, which is typically closed to the public. Families, pulling luggage and holding their children's hands, walked with light steps as SK Telecom (SKT) opened its doors for the 'T Long-Term Customer Forest Retreat Day' for customers who have been with the company for over a decade. The Forest Camp spans approximately 90,000 square meters (about 22 acres) in the Hyangsusan area of Yongin and is not usually accessible to the general public. SKT has partnered exclusively with Everland to grant long-term customers a full day at this expansive space. Participants enjoyed the area as if they had rented it out, with around 50 families attending without feeling crowded. The weather was perfect for an outdoor event, with clear skies and a refreshing breeze. Upon entering, participants chose their spots on the grassy area set up with tents and umbrellas. SKT provided coffee, various beverages, and snacks at the entrance, along with bubble-making tools for families with young children. Equipment for family activities, such as books and board games, was available for rent on-site, and activities like glider-making and frisbee-throwing captured the children's attention. Even those who arrived empty-handed found everything they needed at the event. Children ran around on the grass, launching gliders into the air. The red foam gliders, emblazoned with the SK Telecom logo, soared against the blue sky. Parents laughed along with their children as they cheered. One participant remarked, "My child is busy playing with bubbles, throwing frisbees, and flying gliders without a moment's rest," smiling as they spoke. At 10 a.m., the main forest retreat program began, divided into sessions for children and adults. Although participation was voluntary, nearly all attendees joined in. The morning session quickly filled up, and the reporter signed up for the 1 p.m. session. Lunch consisted of a bento box filled with bulgogi, shrimp, fruits, and a variety of side dishes, providing ample portions. After the meal, participants followed a guide into the woods. This area has been part of the land since the days of the former natural farm that preceded Everland. As they walked along paths lined with pine and oak trees, the guide explained the names and origins of the plants they encountered. The stories behind the plants made the hour-long walk engaging and enjoyable, and participants felt a sense of rejuvenation as they strolled through the dappled sunlight of the forest. Stopping briefly in a clearing, participants picked up bamboo poles as instructed by the guide. They engaged in stretching exercises and simple cooperative games using the poles. The sight of everyone forming a circle to raise the bamboo poles together created a picturesque moment filled with laughter. After the forest retreat walk, families returned to the grassy area for recreational activities. The host led light-hearted games, including quizzes with prizes for winners. During a segment introducing the longest-tenured SKT customer, a woman who has been with the company for 29 years took the spotlight. When sharing her story, one participant took the microphone to express a desire to reconnect with her father, who had called to congratulate her on passing the police exam 15 years ago. This heartfelt moment brought tears to many eyes. The recreation session concluded after activities like children's dance time and an adult breath-holding game, lasting just over an hour. Following the recreation, participants were given free time until the 4 p.m. departure. Many lounged on blankets, reading books or enjoying one last glider flight with their children. Satisfaction was evident on their faces. One female customer attending with her husband and child said, "This is my first time attending, and I am very pleased and satisfied. I truly feel treated as a customer, and I am very grateful for this wonderful venue." Now in its third year, the forest retreat operates in both spring and fall each year. This spring season, from May 3 to 18, a total of 1,800 participants were invited across six sessions. The competition rate soared from 130 to 1 in the first year to 636 to 1 this year. Everyone who attended expressed satisfaction with the experience. SKT plans to expand its offerings for long-term customers, starting with the forest retreat, followed by gourmet events, amusement park invitations, and musical performances. The strategy aims to reward customers who have been with the company for a decade with experiences rather than discounts.* This article has been translated by AI. 2026-05-14 15:42:24
  • AI Security Battle Intensifies Among Major U.S. Tech Firms
    AI Security Battle Intensifies Among Major U.S. Tech Firms On May 13, major international news outlets focused on the AI security front, highlighting OpenAI's new platform, the fallout from Claude Mythos, and efforts to thwart weaponized AI attacks. OpenAI Launches Cybersecurity Platform 'Daybreak' OpenAI has officially entered the cybersecurity market with the launch of its enterprise platform, 'Daybreak.' This platform combines the GPT-5.5 model with OpenAI's proprietary security engine, 'Codex Security,' designed to support everything from software vulnerability detection to patch validation and automation. Unlike traditional security vendor solutions, OpenAI positions Daybreak as a 'control layer for application security infrastructure.' Industry experts view Daybreak as a direct competitor to Anthropic's 'Project Glasswing and Claude Mythos.' Market research firm Future Group assesses that OpenAI aims to establish a governance role above the app security agent layer. Ongoing Fallout from Claude Mythos; EU and U.S. Involvement The impact surrounding Anthropic's cybersecurity-focused model, Claude Mythos, continues to reverberate. While OpenAI has agreed to provide GPT-5.5 cyber access to the EU, Anthropic has yet to finalize its European rollout of Mythos. A month after its launch, the European Commission still has not secured access rights. In the U.S., the situation is urgent. White House Chief of Staff Suzy Wiles, Treasury Secretary Scott Bessen, and National Cyber Director Sean Kerckhove are directly involved in responding to Mythos, with the White House officially opposing Anthropic's plans to expand Mythos access. Consequently, major AI firms like Google and Microsoft are entering governance collaborations, including pre-launch model review agreements with the U.S. Department of Commerce's AI Standards Innovation Center (CAISI). Google Preemptively Blocks AI-Driven Zero-Day Attack Attempts Google's Threat Intelligence Group (GTIG) announced that it has preemptively blocked a 'large-scale exploit operation' where hackers attempted to find and exploit zero-day vulnerabilities using AI models. Google confirmed that its Gemini model was not used in these attacks. This incident marks the official recognition of cyber threats weaponized by AI, indicating that the competition in AI security has escalated beyond mere corporate rivalry to a serious threat response level.* This article has been translated by AI. 2026-05-14 08:58:58
  • National Tax Service to Track Cryptocurrency Wallet Addresses with New System
    National Tax Service to Track Cryptocurrency Wallet Addresses with New System The National Tax Service (NTS) has officially begun developing a dedicated computer system for tracking and analyzing cryptocurrency transactions. On May 11, the NTS announced that the Seoul Regional Tax Office's Information Center and Nanal SMI held an inauguration ceremony for the 'Integrated Analysis System for Virtual Assets' on May 8. The project aims to be completed by the end of December this year, with a total budget of 2.99 billion won (including value-added tax, excluding procurement fees). The NTS's initiative is driven by the structural characteristics of virtual assets. Due to their anonymity and decentralization, cryptocurrencies have increasingly been misused for illegal activities such as money laundering, irregular gifting, and offshore tax evasion. Additionally, starting in 2027, cryptocurrency businesses will be required to submit individual transaction data to the NTS, making it urgent to establish a systematic computerized foundation for utilizing large volumes of transaction information in tax audits. The core of the system will be divided into three main components. First, it will create a foundation for the continuous and systematic integration of cryptocurrency transaction information and blockchain transaction data. The system will periodically collect transaction statements and summaries submitted by cryptocurrency businesses to the NTS, allowing for integrated queries linked with existing tax data for reporting, tax assessments, and audits. For each taxpayer, it will provide information on transaction overviews, changes in virtual asset holdings, and current balances, while enhancing transaction tracking by visualizing identified wallet addresses alongside blockchain transaction data. The system will go beyond simple data collection. It will include features designed to analyze and verify transaction flows suspected of tax evasion, such as money laundering and unreported inheritance or gifting. An analytical framework will also be established to proactively detect unusual transaction patterns and suspicious individuals using AI machine learning and statistical techniques. User convenience and information security have been prioritized in the design principles. The system will automate manual tasks and connect data from various sources while implementing a control system to ensure that transaction information is used only within the necessary scope through access management and log tracking. The NTS stated, "By systematically managing vast amounts of cryptocurrency transaction data, we expect to enhance the efficiency of related operations and strengthen the infrastructure for virtual assets. We aim to contribute to tax fairness and justice by blocking illegal transactions such as money laundering, irregular gifting, and offshore tax evasion, as well as analyzing suspected tax evasion cases."* This article has been translated by AI. 2026-05-12 04:27:39
  • Companies Struggle with AI Transition: Redesigning Operations is Essential
    Companies Struggle with AI Transition: Redesigning Operations is Essential Despite ongoing attempts by South Korean companies to transition to artificial intelligence (AI), many are stalling at the implementation stage and failing to achieve tangible results. Most failures stem from treating AI as a mere tool, highlighting the need for a fundamental redesign of work processes around AI. According to the Korea Artificial Intelligence and Software Industry Association (KOSA), the utilization rate of AI in South Korea's manufacturing sector remains at just 17.9%. The primary reason for the lack of AI adoption is the difficulty in identifying areas and processes suitable for AI implementation, cited by 41.6% of respondents. A recent report from the Super Large AI Promotion Council identified four main reasons for failures in AI transitions: unclear responsibility and roles (R&R), lack of operational monitoring, unaddressed data discrepancies on-site, and governance issues that hinder implementation. While proof of concept (PoC) projects may succeed, unclear responsibilities during actual operations often lead to a lack of trust on the ground, resulting in projects being abandoned. The report succinctly states, "PoC is a snapshot, but validation is a four-season process," meaning that success at one point in time does not guarantee effective operations in changing conditions. Unique factors contributing to failures in the South Korean corporate environment have also been identified. In a case study from a knowledge and office sector company, an engineer noted, "The issue wasn't that AI couldn't write; it was that when data came in HWP format, half of the project shifted from generation to input processing." This highlights the warning that if the initial parsing pipeline is not designed to recognize HWP as a file format rather than a language, operational costs will continue to rise. The predominance of HWP format in public documents presents a significant barrier to AI adoption in South Korean companies. Similar patterns have been observed in the trade and logistics sectors. In a joint project between LabelUp and Team Reboot, an engineer remarked, "Don’t assume PDFs will be neatly formatted text." Effective automation of unstructured documents requires simultaneous design of structuring, validation, and recommendation stages; if the extraction fields are unstable, the reliability of subsequent recommendations collapses. Conversely, companies that have redesigned their operational processes are seeing clear benefits. Company B reported a reduction of up to 73% in full-time equivalent (FTE) staffing after implementing AI agents, while increasing its internal AI productivity index by 35%. LabelUp and Team Reboot reduced document processing time by over 60% on average and achieved a 99.2% accuracy rate in HS Code classification during final validation, which included human review loops. In terms of optimizing AI infrastructure, companies have cut GPU training costs by approximately 80% compared to AWS on-demand services, tripling development productivity and increasing GPU utilization from 20% to over 85%. Analysis indicates that the criteria for success in AI transitions hinge not on model accuracy but on the completeness of operational design post-deployment. Key elements such as monitoring systems, incident response procedures, user training, and standardization of R&R are essential for translating efforts into tangible business outcomes. Successful companies have commonly applied five design patterns: task decomposition and agent delegation, assetization of unstructured data, real-time edge analysis, optimization, and the integration of security and governance. The Super Large AI Promotion Council concluded that the true determinants of success in AI transitions lie not in the initial implementation but in the subsequent operational design, accountability structures, and governance frameworks.* This article has been translated by AI. 2026-05-12 04:17:42
  • Public Frustration Grows Over Fuel Prices and Telecom Fees Amid Corporate Profits
    Public Frustration Grows Over Fuel Prices and Telecom Fees Amid Corporate Profits In January 2011, former President Lee Myung-bak sparked a stir in the oil industry when he remarked on the disparity between international oil prices and domestic fuel costs, saying, "Fuel prices are strange." His comments led to government inspections of gas stations and intense media scrutiny of oil company profits, igniting public outrage over rising fuel costs. Fifteen years later, amid soaring international oil prices due to the war in Iran, President Lee Jae-myung convened an emergency cabinet meeting, directing the establishment of a maximum price for gas stations. He expressed strong dissatisfaction, noting, "Prices vary from morning to noon to evening, and some stations have raised prices by nearly 200 won per liter in a single day," pointing to unethical pricing practices by certain gas stations. According to officials from the Ministry of Trade, Industry and Energy, the price inspections that began at gas stations are now extending to oil companies. Due to the surge in oil prices linked to the Iran conflict, the first-quarter operating profits of major refiners, including SK Innovation, S-Oil, GS Caltex, and HD Hyundai Oilbank, are projected to approach 5 trillion won. Some lawmakers are reportedly preparing to pressure these companies to return their profits. When international oil prices rise, the selling price of petroleum products made from already secured crude oil increases, leading to a significant short-term boost in operating profits. Notably, over 80% of the operating profits of domestic refiners come from exports, while the margin that refiners earn from domestic gas station sales is only a few dozen won per liter after taxes. Telecom companies are facing similar scrutiny. The first-quarter results of the three major telecom firms are expected to exceed market expectations. Industry insiders attribute this resilience not to wireless communications but to increased demand for cloud services and data centers, despite subscriber losses due to hacking incidents. The wireless communication sector itself has stagnated due to the government's stringent policies against significant rate hikes. Current telecom companies have found new revenue sources through investments in AI infrastructure rather than extracting profits from consumers. Nevertheless, whenever the economy falters, the government consistently targets telecom companies. The Lee Myung-bak administration promised a 20% reduction in communication fees, while the Park Geun-hye administration called for "half-price telecom fees." The Moon Jae-in administration sought to eliminate basic fees and increase discount rates, and the Yoon Suk-yeol administration has made expanding low-cost 5G plans a priority. None of these administrations have succeeded in achieving substantial rate reductions. Under the current government, the Ministry of Science and ICT is discussing improvements to telecom plans with the three major companies, which are expected to launch 5G plans priced in the 20,000 won range this year. Industry insiders express skepticism, noting that the government continues to seek solutions to economic stimulation through telecom fees, just as previous administrations did. By creating villains in the oil and telecom sectors, the government has used them as scapegoats to address public frustration stemming from economic downturns. It is now time to focus on the real issues at hand. Labor unions in large corporations, which pay salaries exceeding 100 million won, are demanding performance bonuses in the millions of won and have initiated strikes. In contrast, many workers in small and medium-sized enterprises, which make up a significant portion of the labor force, have received notices of wage freezes due to the economic downturn. As large corporations absorb the pressure of rising labor costs by lowering supplier prices, the burden will undoubtedly fall on subcontractors and their employees. Many workers will tighten their budgets, and it will become increasingly common to see neighborhood restaurants changing their signs every few months. As lunch prices soar above 10,000 won, images of workers demanding performance bonuses of 600 million won make headlines. Today, the government continues to target easy scapegoats while proposing economic stimulus measures. The entities generating billions in operating profits are not the oil companies or telecom firms profiting from cloud services, but rather a troubling structure that mocks the entire population while masquerading as the underprivileged.* This article has been translated by AI. 2026-05-08 16:52:17
  • AI Big Three Cut New-Model Release Cycles to About 50 Days, Raising Barriers for South Korea
    AI Big Three Cut New-Model Release Cycles to About 50 Days, Raising Barriers for South Korea Google, OpenAI and Anthropic have effectively compressed their new AI model release cycles to roughly a monthly cadence, as competition accelerates in the era of agent AI, where falling behind is widely seen as fatal. An analysis of the three companies’ major model launches over the past six months shows an average interval of 50 days, down about 62% from the three-year average of 130 days. By company, Google posted the shortest average at 39 days, followed by OpenAI at 43 days and Anthropic at 68 days. OpenAI unveiled GPT-5.5 on April 23, 49 days after releasing GPT-5.4. Its recent sequence — GPT-5 (August 2025), GPT-5.1 (November), GPT-5.2 (December), GPT-5.3 (March), GPT-5.4 (March) and GPT-5.5 (April) — contrasts with the roughly eight months it took to move from GPT-4 to GPT-4 Turbo. Anthropic released Claude Opus 4.7 on April 16, 70 days after Opus 4.6. The company said Opus 4.7 delivered meaningful gains over Opus 4.6 in coding, vision and self-verification. Google has maintained the fastest cadence, rolling out Gemini 3 in December, Gemini 3 Pro in January, Gemini 3.1 Pro in February, 3.1 Flash Lite in March and Gemma 4 in April. The company has kept its pattern of one major version per year while using frequent sub-versions to expand practical market share. The acceleration is being driven by three factors, the article said. First is the full-scale shift to agent AI: as use expands beyond chat into coding, data analysis and workplace automation, the minimum performance bar is rising quickly, and laggards can see immediate customer churn. Second is a sharp drop in inference costs. Flagship model operating costs constrained release speed as recently as two years ago, but per-token prices have fallen rapidly, easing that limit. Third is an industry practice of branding improvements in fine-tuning, safety and instruction-following as “new models,” favoring fast, incremental updates over major architectural overhauls. Competition between OpenAI and Anthropic is also heating up amid financial pressures. With both aiming for initial public offerings in the second half of this year, investors are focused on which company has the strongest model, and the market mood is increasingly winner-take-all. Investment banking circles view profitability for the No. 2 player as far below the leader’s, adding pressure to secure a performance edge before IPOs — and further shortening release cycles. Against that backdrop, concerns are growing that South Korea’s independent foundation models will see their international competitiveness erode. Government-led support programs can take months just to select operators. The National Growth Fund’s plan to invest 560 billion won in Upstage’s independent foundation model is a meaningful attempt, but questions remain about whether Korean models can break into global markets as the big three push out new releases about every 50 days.* This article has been translated by AI. 2026-05-05 17:00:15
  • Korea Aerospace Administration chief meets U.S. space firms after satellite launch
    Korea Aerospace Administration chief meets U.S. space firms after satellite launch Oh Tae-seok, administrator of the Korea Aerospace Administration, met with major U.S. space companies in the United States after the successful launch of South Korea’s Next-Generation Mid-Sized Satellite No. 2, seeking cooperation to strengthen the competitiveness of the country’s private space industry. The agency said Tuesday that Oh on May 3 (local time) successfully completed his duties as head of the launch management team at Vandenberg Space Force Base. The satellite was launched at 4 p.m. Korea time aboard a Falcon 9 rocket and, about 60 minutes later, was placed into a 498-kilometer sun-synchronous orbit. At 5:15 p.m. the same day, communications with the Svalbard ground station in Norway confirmed normal operation of the spacecraft and key systems. The agency described it as the first mid-sized satellite independently developed by a private company based on 500-kilogram-class standard platform technology, calling it a symbolic result showing a shift in space development from government-led to private-led efforts. Oh continued what the agency described as on-site outreach the next day, May 4, visiting key space companies near California. He first went to Umbra Space to review the company’s development and use of synthetic aperture radar-based small satellites. With Umbra Space, which has been rapidly expanding commercial satellite data services for disaster response and security, he discussed cooperation on private-led small-satellite development and ways to use satellite data. Oh then met with SpaceX officials and formally requested coordination on the schedule for Next-Generation Mid-Sized Satellite No. 4, which is set for launch in the second half of the year. The agency said the meeting focused on schedule management after a previous delay, when a planned joint launch of satellites No. 2 and No. 4 was changed to separate launches due to SpaceX circumstances. The sides also discussed launch plans for follow-on satellites the agency is pursuing, including GEO-KOMPSAT-3, and exchanged views on the global launch market and the direction of technological development. “Building on the successful launch of the private-led Next-Generation Mid-Sized Satellite No. 2, Korea’s space industry also needs a rapid transition to a private-led approach,” Oh said. He said the agency would actively pursue expanded cooperation with major space powers and companies to strengthen the space-technology competitiveness of South Korean private firms. * This article has been translated by AI. 2026-05-05 13:52:12