East Asia

South Korea

South Korea’s public technical-safety ecosystem is anchored by Korea AISI plus university evaluation work at Soongsil and method-level safety research at KAIST.

Government posture

Korea AISI publicly emphasizes dangerous-capability categories, evaluation infrastructure, international cooperation, and ETRI affiliation.

Standards and evaluation

The clearest public evaluation signals are Korea AISI’s risk and infrastructure pages, its 42-model disclosure, and Soongsil’s 20-model benchmark with 57 attacks.

Research ecosystem

Public work covers jailbreak defense, transformer interpretability, prompt-injection and jailbreak evaluation, and broader security framing for Korean and foreign foundation models.

International engagement

Official AISI materials frame Korea’s institute as part of the wider international AISI network.

Seed entities

AI Safety Institute Korea

인공지능안전연구소

South Korea

Ai Safety InstituteTier 1

Major Korean public evaluation node with explicit risk categories, benchmark infrastructure, and public model-evaluation activity.

Frontier Ai SafetyEvaluationTesting InfrastructurePolicy Regulation

Confidence: HighVerified: 2026-06-06

Ministry of Science and ICT

과학기술정보통신부

South Korea

GovernmentTier 1

The ministry that launched Korea AISI and publicly ties it to post-Seoul-Summit implementation and international AI safety networking.

Ai GovernancePolicy RegulationInternational Dialogue

Confidence: HighVerified: 2026-06-05

Electronics and Telecommunications Research Institute

한국전자통신연구원

South Korea

Research GroupTier 1Needs primary

The clearest public institutional home for Korea AISI in reviewed sources.

Technical Safety ResearchEvaluationTesting Infrastructure

Confidence: HighVerified: 2026-06-05

Korea AI Safety Consortium

AI안전컨소시엄

South Korea

Industry AllianceTier 1

Consortium formed at Korea AISI’s launch to support framework work, subgroup collaboration, and joint-testing discussions.

Capacity BuildingEvaluationInternational DialogueAi Governance

Confidence: HighVerified: 2026-06-05

Soongsil University AI Safety Research Center

숭실대학교 AI안전성연구센터

South Korea

Research GroupTier 1

University-based Korean group with direct public evidence of large-model security and safety benchmarking using prompt-injection and jailbreak attacks.

EvaluationRed TeamingModel SecurityCybersecurity

Confidence: HighVerified: 2026-06-06

KAIST Artificial Intelligence and Probabilistic Reasoning Lab

KAIST AIPR Lab

South Korea

Research GroupTier 1

KAIST lab with direct public work on jailbreak defense for misaligned instructions and transformer interpretability.

Technical AlignmentRed TeamingModel Security

Confidence: HighVerified: 2026-06-06

Open questions

Need public links to Korea AISI datasets and tools, not just institutional descriptions.
Need stronger direct-source evidence for additional Korean university labs beyond KAIST AIPR and Soongsil.