Korea's housing squeeze deepens as justice reform and AI move into practice

Housing policy met a more expensive money environment

The week opened with two pressures that sit outside housing policy but shape what households can afford. The won closed at 1,385.5 per dollar on September 19 after the US Federal Reserve raised its policy rate by 0.25 percentage point, while elevated oil prices kept import costs in focus. The weaker currency and expensive energy increased the risk that fuel, transport and production costs would remain a burden even before mortgage payments were considered.

Against that backdrop, the government announced one of the largest housing-supply commitments of the year. Its 1.19 million-home public-sector plan through 2030 puts roughly 920,000 homes, or 77 percent of the total, in the capital region. The plan also aims to expand the public-rental stock and open part of the system more widely to young and middle-income households.

Supply, however, is only one side of affordability. A Korea Research Institute for Human Settlements study covering 247 months found that a 0.25 percentage-point policy-rate increase was associated with a 1.2 percent decline in nationwide apartment prices six months later during rising-rate cycles. The same week, separate reporting estimated that a quarter-point increase could add about 7 trillion won to annual interest costs across borrowers.

Those effects do not move every neighborhood in the same direction. Seoul apartment prices rose for an 85th consecutive week, matching the longest run in the available Korea Real Estate Board series, but the citywide increase hid a split market. Gangnam, Seocho and Songpa fell while several northern Seoul districts rose, and parts of Gyeonggi connected to the semiconductor corridor increased even faster.

Debt indicators make that combination harder to manage. Bank-loan delinquency rose during the week, the Bank of Korea warned that high rates could push more borrowers into arrears next spring, and separate data showed a growing share of ultra-high-risk borrowers. A large supply target can improve the medium-term market, but households encounter housing policy through monthly payments and available listings long before a 2030 construction target is complete.

Prosecution reform reached the point where staffing numbers became policy

The prosecution overhaul became more concrete over two consecutive days. A Justice Ministry draft initially described a system that could retain the existing statutory headcount of 2,292 prosecutors while changing senior posts and shifting the organization toward prosecution and case review. The following day, Cabinet-approved launch arrangements set the new Prosecution Service to begin on October 2 with about 1,900 prosecutors, roughly 80 to 85 percent of the current statutory ceiling.

The change between the draft and launch plan showed what implementation looks like in practice. The government is separating investigative functions from prosecution while also deciding how many prosecutors and investigators the new institution needs. Officials have said further statutory reductions can follow after workload analysis. The immediate test is whether cases continue to move while agencies adjust to new responsibilities.

The institutional transition was happening alongside a separate dispute over a Supreme Court vacancy. Chief Justice Cho Hee-dae rejected a renewed request to recommend a different candidate, while the presidential office argued that the president has substantive appointment authority. By late week the disagreement had lasted more than 200 days, turning an individual vacancy into a continuing argument over the boundaries between nomination and appointment powers.

Courts were also still dealing with the consequences of the December 3 martial-law episode. On September 22, five former senior military commanders received prison terms ranging from nine to 18 years for their roles in carrying out the operation. That case is legally separate from the prosecution overhaul, but it underscored how much institutional change and accountability work the justice system is handling at the same time.

The October 2 launch will provide the first operational evidence. Headcount alone cannot show whether the redesign works. Case-transfer delays, review workloads, staffing gaps and the division of responsibility between investigators and prosecutors will matter more once the new structure starts handling ordinary cases.

AI moved from an industrial strategy into ordinary systems and ordinary disputes

South Korea's AI policy is increasingly visible outside research labs and data centers. The government-backed AI for Everyone project is preparing an October beta ahead of a formal December launch. Consortiums led by SK Telecom, Kakao and KT are developing general chatbots and task-oriented agents intended to work through familiar consumer channels, including services that could help with applications, reservations and payments.

At the workplace level, automation is already becoming part of hiring and evaluation. In one survey of 396 companies, 86 said AI tools were officially used in recruitment. AI is also entering employee and executive assessments, including measures of adoption, output and in some cases the cost of model use. Korean law gives applicants a right to request an explanation of an AI-supported hiring decision, but that protection is difficult to use when applicants are not clearly told that automation influenced the process.

The same expansion is creating a second set of rules around misuse and synthetic content. Police reported a sharp increase in undercover investigations involving deepfake crimes. New watermark penalties are aimed primarily at service providers, leaving questions about how individual users are treated. Other reporting described AI-generated receipts and shipping documents being used in secondhand-trading fraud, while AI-enabled glasses exposed gaps in exam and privacy rules.

Industrial adoption continued in parallel. Korean manufacturers reported measurable productivity gains from factory AI, companies expanded data-center cooling and power partnerships, and robots were moving into routine inspection and service work. These are different uses with different risks, so one regulatory answer will not fit all of them.

The October public-service beta will be a useful test because it combines scale, accessibility and action. A chatbot that only answers questions creates one kind of risk. An agent that submits an application or makes a payment creates another. Reliability, disclosure and the ability to correct mistakes become part of the public-service design rather than optional technical features.

Two security cases tested how much information should become public, and when

The transfer of two North Korean soldiers captured in Ukraine became a dispute over confidentiality after the men arrived in South Korea. President Lee Jae Myung said officials had kept the transfer quiet because disclosure could affect lives, national security and diplomacy. Ukraine's public confirmation changed that arrangement and prompted criticism from Seoul. Opposition politicians, meanwhile, questioned whether a major security development had been kept from the public for too long.

The important boundary remains the soldiers themselves. Public reporting does not establish their wishes, legal status or current debriefing process. Those missing facts limit how confidently the transfer can be interpreted and make protection a practical concern rather than an abstract argument about transparency.

A different information problem developed around the DMZ mine blast that injured three South Korean officers. Critics questioned the pace of the on-site investigation and warned that delay could affect evidence. The government responded that investigators were dealing with a potentially mined area and had to secure access before sending more personnel into danger. The available evidence did not establish a cover-up or deliberate manipulation, and the origin of the explosive device remained under investigation.

These were distinct cases, but each forced officials to make a timing decision about disclosure under conditions of risk. Public accountability still requires facts, yet releasing operational details too early can create its own consequences. In both cases, the strongest next step is factual: the soldiers' status and wishes where disclosure is safe, and the forensic findings from the DMZ site once investigators can establish them.

Chuseok concentrated household costs, travel demand and a predictable safety burden

Holiday policy moved from preparation into execution. The government waived highway tolls for four days, added rail capacity and extended food discounts as millions of people traveled. Earlier planning estimated that 12.99 million people could move on Chuseok day, while the shorter break was expected to compress traffic more tightly than in some previous years.

Household data showed why the holiday remained a cost issue even after discount programs were announced. Reporting during the week put typical holiday spending near 730,000 won while the share of workers receiving Chuseok bonuses declined. Convenience-store delivery demand also jumped during the logistics gap, showing how quickly spending patterns adapt when regular delivery and retail schedules change.

The most serious holiday statistic concerned safety inside homes. Police data showed that during the previous seven-day Chuseok holiday, 112 received 8,017 domestic-violence reports, or 1,145 reports a day. That was 44.5 percent above the ordinary daily average of 792. The holiday daily figure had also risen from 914 in 2021 to 1,145 in the latest comparison.

The numbers do not identify one cause for domestic violence, but they do identify a recurring period of elevated demand for police and victim-support services. That makes Chuseok staffing and protection planning measurable. The holiday is predictable, and so is the need to make sure urgent cases are not lost inside the larger call volume.