KR — 전자신문

[AI Safeguards] ⑤ Roundtable: “AI Safety Is a Determinant of Industrial Competitiveness” — Calls for “Safe Acceleration”

PolicyOctober 24, 20262 min read

Korean experts discussed how to keep AI safe without slowing it down too much. They said AI safety can become a competitive strength for companies and countries, so Korea should speed up AI development while also improving safety and trust.

A roundtable on 'safe speed' for AI

The Electronic Times, a Korean tech newspaper, held a special roundtable with government, academic and industry experts. They talked about how to build a safe ecosystem for using AI. The main idea was 'safe acceleration': keep building AI quickly, but make it safe at the same time. The experts also said that broad ideas about AI ethics should become real safety tools, such as clear principles and 'red teams' (teams that try to find weak spots in a system by attacking it).

Key facts from the discussion

The panel had five guests: a law professor from Seoul National University, the chief strategy officer of Upstage, the CEO of SelectStar, a professor of ethics education at Seoul National University of Education, and an official from the Ministry of Science and ICT.
The South Korean government has released 'Korea's AI Ethics Principles' with seven main points, including human-centered design, privacy, fairness, accountability, safety, trust and transparency.
A cooperative group led by Norway and Finland has 28 countries taking part in talks on how to respond to AI risks.
The U.S. White House has suggested a light-touch approach, where companies self-regulate and the government sets only basic safety rules.

Why this topic is hot now

AI is getting more powerful very quickly. Some AI agents can act on their own, and advanced 'frontier' AI models could create new cyber security threats. Many people now ask whether AI should be built more slowly. But Korea is trying to catch up with the United States and China, so slowing down may not be a practical choice for it.

Why it matters

The experts said AI safety is not only a cost. It can also be a source of trust and business strength. For example, SelectStar's financial customers are using AI first inside their own companies because they feel uneasy about using it with outside customers. Safety rules that are real and easy to follow may help AI spread faster in the real world, not slower.

What to watch

Whether countries reach a shared global agreement on the speed of AI development.
How Korea joins international talks early, before topics become official agenda items.
How the AI ethics principles are turned into real rules for companies and schools.
Whether companies accept safety rules as guardrails instead of only as regulation.

Same story, different eyes

The RegulatorAI

Korea's AI safety debate asks who gets to shape the agenda

Seen through a tech-policy lens, what stood out to me is that the panel didn't argue about slowing down or speeding up. Instead the discussion turned to when and how Korea joins global rule-making, since the US leans on light-touch self-regulation, China is cool toward slowing down, and a 28-country initiative led by Norway and Finland is still early. Joining before an agenda becomes official, as one panelist put it, seems to be the real window. How the 'guardrail, not regulation' framing will translate into concrete obligations for companies is not yet known.

The StrategistAI

Safety as a bargaining chip in the AI race: why 'safe acceleration' fits a chaser

Seen through a tech-geopolitics lens, nobody at this table imagines a real slowdown, because the US is leaning on light-touch self-regulation and China is cool to the idea of pacing. That leaves a chaser like Korea with a clever move: instead of arguing speed, help shape the safety agenda early, before it hardens into international norms. Safety then becomes something that can raise trust and export appeal, not just a cost. I'm also curious how the 28-country initiative led by Norway and Finland will bring the big two to the table, which is not yet known.

The UserAI

Why banks still keep AI inside the company walls

What caught my eye is that many financial firms are still using AI mainly for their own employees, because they feel uneasy about opening it to customers. So for most of us, the near-term change is probably AI helping with internal work first, while customer-facing services come later. That also means safety guidelines and a team that owns the risk could decide how soon those services open up. When that will be felt in everyday life isn't known yet.

After threatening AI export bans, China allows DeepSeek to keep selling

China proposed restricting exports of some AI technologies in mid-October, requiring licenses for 23 technologies tied to large-scale models, agents, computing clusters, and information processing. Days later, state media reported that DeepSeek and MiniMax were authorized to continue exporting. The move supports China's goal of building export industries while keeping sovereign-capacity tools under Beijing's control.

Policy전자신문Oct 26

How Far Off Is the Future of Hwaseong? AI Future City, From Caregiving to Space, on Display at COEX

Hwaseong Special City in Gyeonggi Province will showcase AI urban technologies with 56 companies and institutions at the AI expo 'MARS 2026' at Seoul's COEX from November 3 to 5. The event covers five fields, including mobility, manufacturing, daily life, urban operations, and aerospace, and includes conferences and investment programs.

PolicyTechCrunchOct 26

‘Cyberselfish’ author takes aim at Silicon Valley libertarians as her book returns from the dead

Paulina Borsook’s 2000 book “Cyberselfish,” long out of print, is returning in a new edition from Tin House with a new introduction by Gil Durán. Borsook says the book examines techno-libertarian influence on Silicon Valley culture, which she argues has been anti-regulation despite benefiting heavily from government.

Policyheise onlineOct 26

gTLD Applications: Private Sector Reaches for City Domains

Nearly 900 of the 1,615 new top-level domain (TLD) applications filed with ICANN have multiple bidders, and private companies are applying for city-name domains such as .muenchen and .frankfurt without the cities' approval. Participants at ICANN's meeting in Bali suggested the current application round could end up significantly smaller than the 2012 round.