South Korea reopens a military-service route for AI researchers
Seoul reopens a path it closed in 2013 to keep AI researchers inside its biggest labs — and the first intake is already undersubscribed.
South Korea will let AI researchers satisfy their military service obligation inside large corporate laboratories from 2027, reviving a route the government shut off in 2013. Eligible science and engineering researchers with master's or doctoral degrees serve three years of research instead of conventional service — the same structure that never stopped at universities, public institutes and smaller companies, but which large businesses lost access to for 14 years. The government has set aside 240 places for master's-level AI researchers in 2027: 120 at large-company laboratories and 120 at government-funded and defense-industry research institutes, with the corporate allocations running three years through 2029. Ministry of Science and ICT data obtained by Rep. Hwang Jung-a of the ruling Democratic Party lists LG Electronics, Samsung Electronics, Samsung Medison and HD Hyundai Robotics as the four companies recommended to take part.
The demand side is the interesting part. Those four companies, across nine research institutes, requested 81 placements — just over two-thirds of the 120 reserved for large-company labs, leaving 39 unfilled at the recommendation stage. Qualification is narrow: a company must spend at least 5 percent of revenue on R&D, work in a field that does not directly compete with smaller businesses, and demonstrate objectively verifiable research achievements. Industry officials told the Korea Herald that some applicants failed to secure recommendations on those conditions, and the Military Manpower Administration still has to approve the institutes and their allocations, so the final number could land below 81.
Seoul's motive is retention, and the arithmetic is unflattering to it. Korea produces roughly 80 percent of the world's high-bandwidth memory, and its AI buildout has already bent the domestic labour market — we covered the central bank's count of how Korea lost 285,000 youth jobs in AI-exposed sectors between June 2022 and June 2026, and the mechanism behind it in AI is deleting the first rung, not the senior job. Against that, a draft carve-out that fills two-thirds of its slots is a rounding error, and Hwang's own conclusion is that rigid eligibility is what caps it. Three years of lab research is also a long commitment in a field that re-specializes every few months.
The Lunduke Journal says AI-written code accounted for 17.25 percent of all Linux kernel patches submitted in September, after 1,634 AI-authored submissions landed in a single week. Treat that as a claim, not a measurement. The kernel has no standard field for declaring AI assistance, so the figure comes from one journalist's tally of disclosure patterns rather than from maintainer data, and the outlets repeating it have not verified the underlying method. The direction is not in dispute. Torvalds and the maintainers have finalized a policy that permits AI-assisted patches while placing full review liability on the human submitter, and the opposing camp is still fighting it out: we covered Debian opens project-wide vote on banning LLM contributions, a ballot that has since run past its own deadline with the project split across eight overlapping options. Both halves point the same way — the argument is no longer whether AI code enters the kernel, but who signs for it.
What to watch: whether the Military Manpower Administration approves all four labs at full allocation, and whether anyone outside the Lunduke Journal reproduces the kernel tally.
Korea is offering a draft exemption to keep its AI researchers home — does a talent pipeline that needs legal carve-outs say more about the market or the policy? Tell us in the comments.
