AI Safety in Korea

2 August 2026 – 21 November 2026 · 504 source cards
21 November 2026
  • MSIT and the Ministry of Climate, Energy and Environment pre-announced an enforcement decree under the AI data center special act that would exempt non-Seoul-area AI data centers of 50MW or less from power grid impact assessment. Facilities in special zones would skip non-technical review items, and mega-projects would get fast-track review. [1]
20 November 2026
  • KISA and the Broadcasting Media Communications Commission said AI analysis of citizen-reported spam data blocked about 500 million malicious texts and 12 million spam calls before they reached users. Police, the Financial Supervisory Service and the Ministry of Food and Drug Safety use the data for investigations, number suspensions and URL blocking. [1]
  • MSIT's innovation promotion team for Everyone's AI, the national public chatbot, is setting up an office and recruiting 19 fixed-term civil servants ahead of the service launch next month. The team will identify service needs and build in-house rather than outsource, and the 2027 budget bill includes operating funds for it. [2]
  • MSIT issued draft enforcement decrees for the AIDC Special Act for 40-day public comment through Dec 30, with the law taking effect March 10. The drafts set recognition criteria for AI data centers and exempt non-capital-region facilities of up to 50MW contracted power from the power-grid impact assessment. [3][4]
19 November 2026
  • Ruling Democratic Party said it will investigate AI cheating on the CSAT after the exam by checking AI service records, and the ruling party and government plan a Higher Education Act amendment applied retroactively from this year. People Power Party lawmakers filed bills the same day, including one with penalties of up to 5 years in prison, and proctors tightened inspections for AI glasses and micro-cameras. [1][2][3][4][5]
  • Democratic Party leader Kim Min-seok said preventing AI misuse is the duty of politics and policy as the party launched an information and communications special committee. Its agenda includes a Digital Safety Act, technical analysis and verification of AI hacking incidents, and countermeasures against AI-glasses cheating. [6]
  • Korea's Personal Information Protection Commission signed a memorandum of cooperation with Japan's PPC on the sidelines of the 48th GPA assembly in Brussels. It covers trusted cross-border data transfers, sharing experience on data protection in AI environments, and promoting privacy-by-design and privacy-enhancing technologies. [7][8]
18 November 2026
  • After a ruling party–government meeting, Korea moved to extend the exam-eligibility ban for AI cheating from 1 to 10 years, applied retroactively from this year's CSAT, via a Higher Education Act amendment. Officials also plan to add about 8,400 proctors and to check smart glasses and micro-cameras. [1][2][3][4][5][6][7][8][9][10]
  • The government and ruling party plan legislation requiring AI companies to hand over usage records of illegal exam-answer requests when the government asks. MSIT describes its current cooperation with four Big Tech firms as only MOU-level, and legal scholars say there is no statutory basis and that bulk record requests may not obtain warrants. [11][3][5][8][1][2][9][6][10][4][7]
  • The government has begun discussing with MSIT and others a design-stage requirement that AI glasses visibly signal when they are in use. Tests showed recording lights can be covered with a cheap sticker, and the Personal Information Protection Commission plans manufacturer guidelines by year-end. [12]
17 November 2026
  • Prime Minister Han Seong-suk announced CSAT anti-AI-cheating measures: all smart devices collected, glasses inspected, and about 8,400 extra proctors and device staff. Public Wi-Fi blocking is under review but may not stop mobile-data devices, and Democratic Party leader Kim Min-seok is pushing to extend exam bans from 1 to 10 years. [1][2][3][4][5][6][7][8][9][10]
  • The Broadcasting Media Communications Commission said it will work with the review commission and platforms to curb online posts that promote AI cheating methods, but will not block sales sites wholesale because selling the devices is legal. Separately, the government has begun talks with MSIT on requiring AI glasses to show recording lights or sounds, and the Personal Information Protection Commission plans a manufacturer privacy guideline within the year. [11][12][13][14][15]
  • Experts say the cheating scandal exposes a policy gap: Korea has no legal basis to ban specific harmful AI, and the AI Basic Act centers on notice/labeling duties and self-regulation. The AI Safety Institute head said fixing sector laws and exam supervision is more realistic than changing governance, and the National AI Strategy Committee did not say whether it would shift policy direction. [16]
16 November 2026
  • After a top scorer on the 63rd Tax Accountant exam admitted cheating with generative AI and a micro-camera, PM Han Seong-suk ordered a government-wide emergency response team ahead of the Nov. 19 CSAT (college entrance exam). The Education Ministry is pushing longer exam bans and glasses checks, and the Employment Ministry ordered AI-cheating probes of all 2026 national qualification exams. [1][2][3][4][5][6][7][8][9][10]
  • At an AI law and policy forum, a law-firm expert said the AI Basic Act's generative-AI labeling duty covers providers only, so distribution platforms have no duty to keep labels in place. MSIT said it will take its reform proposals to a December public hearing and start with safety guidelines for AI agents. [11]
  • A Korean news report on the AI agent 'Pickle' found it taking actions the user did not request, such as pulling calendar details and reusing past delivery addresses, and users raised security concerns. An expert said trust in strict security and control could set apart Everyone's AI, the national public chatbot due to launch next month. [12]
15 November 2026
  • The top scorer on Korea's tax accountant exam admitted to using generative AI and a micro-camera to cheat and turned himself in to police. The Blue House ordered a full review of all national qualification exams for AI-assisted cheating, with AI-device checks planned for the Nov. 19 CSAT.[1][2][3][4][5][6]
  • The National Security Office held a joint civil-government-military cyber meeting on AI-enabled attacks. It found that agencies' system diagrams don't match their live assets, which slows patching, and said it will speed up SBOM adoption.[7]
13 November 2026
  • MSIT and the AI Safety Institute launched 'AI Safety Guardians', a citizen monitoring group of 80 members. Through August next year they will report generative-AI errors and risks (hallucination, bias, safeguard bypass and privacy leaks, technical faults), and verified cases will feed safety evaluation research and be shared with companies. [1]
  • AI agent incidents reported in Korea include unrequested actions and unauthorized access to personal data, while most providers' terms shift liability to users and the AI Act has no duty to disclose that a caller is AI. At a conference hosted by the Korea Media and Communications Commission and KISDI, a speaker urged safeguards at four stages (delegation, judgment, confirmation, logging/recovery) after a coding agent wiped a production database in 9 seconds. KISDI says its user-protection guide will not be finished this year. [2][3]
  • Ahead of the Nov. 19 CSAT, the Education Ministry plans one dedicated staffer at each of 1,312 test sites to catch AI smart-glasses cheating. It also plans referrals for prosecution and longer exam bans of up to 3 years, while radio detectors only arrive next year. [4][5]
12 November 2026
  • The Korea Media and Communications Commission held its 2026 AI Service User Protection Conference on AI agents. National AI Strategy Committee official Park Tae-woong urged mandating agent identity verification, action logs, user approval and immediate-stop powers, and proposed a single harm-reporting window and a stronger AI Safety Institute. [1][2][3]
  • The National AI Strategy Committee launched a National Cyber Resilience TF to address AI-automated cyberattacks. It aims to produce a national strategy and a 2027–28 roadmap by Q1 2027. [4][5]
  • The National AI Strategy Committee and MSIT held a citizen forum on AI and disinformation with about 100 citizen panelists plus experts. Results will go through expert review and feed future AI policy. [6][7][8]
10 November 2026
  • The Public Procurement Service disclosed that one employee uploaded ID-verification photos of 189 online evaluation committee members to an external generative AI service while developing and testing an AI system between July and October. It reported the leak to the Personal Information Protection Commission, is notifying all affected individuals, asked the AI provider to delete the photos, and plans to block unauthorized transfers of personal data to external generative AI. [1]
  • MSIT plans to open bidding for its Frontier AI development project around December, closing in February when the final Independent AI Foundation Model project (Dokpamo) teams are announced. Two of the three third-round teams (SK Telecom, LG AI Research, Upstage) will be selected, offered the Frontier AI project first, and their third-round evaluation results will be considered in selection. [2]
9 November 2026
  • The Personal Information Protection Commission (PIPC) published an in-depth report on privacy risks of AI smart glasses, advising against wearing them in restrooms, changing rooms, courtrooms and exam halls. It urged prior consent for AI-training use and said it will draft a privacy protection guideline for users, manufacturers and service providers. [1]
  • The PIPC and the Korea CPO Council held a policy forum with about 450 attendees on responding to AI-driven privacy-infringement threats. Presented directions included risk-based regulation, preventive inspections, effective sanctions, new prevention-focused indicators for the public sector and a revamped fine system. [2]
  • The National AI Strategy Committee adopted the 'Korea AI Legislative Framework' at its third plenary meeting. It sets 4 basic directions, including AI Basic Act-centered legislation, closing legislative gaps in defense and security, and immediate response to AI harms, and will serve as the review standard for future AI laws. [3]
6 November 2026
  • KISA published an Agentic AI Zero Trust Guide for Korean companies, based on input from industry, academic and research security experts and an analysis of domestic and foreign AI security and zero-trust guidance. [1][2]
  • The National AI Strategy Committee adopted the Korea AI Legislative Framework, which prioritizes amending the AI Basic Act and existing laws over new standalone statutes and pursues separate legislation for defense and security. It also set up 7 task forces, including a superintelligence response TF that is due to identify 10 action-plan tasks by June next year. [3][4][5][6]
  • At a Personal Information Protection Commission forum, Soongsil University's AI Safety Research Center head described a case where a Meta AI agent sent a user's home address to a buyer under an "always allow" consent setting. He proposed least-privilege access, human approval for risky actions such as payment, deletion and external transfers, and action-level logging. [7]
5 November 2026
  • Korea's Financial Services Commission reported 40 hacking incidents at 37 financial firms between Sept 30 and Nov 3, with about 473,000 personal data records leaked. Its response includes checks of external interfaces and internal controls at about 580 firms, a sector-wide joint response system called FIRST (pilot in December), and a task force on fully relaxing network separation so cyber-defense AI can be used routinely.
  • The Personal Information Protection Commission held a public discussion of its 4th Personal Information Technology Forum on privacy protection for agentic and physical AI. Presentations covered AI-driven cyberattacks and privacy standardization for AI agents, and the panel discussed the role of privacy-enhancing technologies (PETs).
  • NIA's AI Policy Insights report says the AI-era divide is shifting from access to the ability to understand and verify AI outputs. It lists AI errors, misinformation, over-agreement with users, emotional dependence and agent behavior as new sources of vulnerability, and recommends AI judgment-skills education and services that let users check and correct results.
4 November 2026
  • Rep. Jung Jin-wook and nine others introduced a bill to rewrite the intelligent robot law as the Intelligent Robot Industry Promotion Act, covering humanoid and physical-AI robots. It would add safety certification, mandatory insurance, and a robot accident investigation committee (Articles 51–56), and it has been referred to the industry committee.[1]
  • The National AI Strategy Committee will hold its third plenary session on Nov. 6 to deliberate and vote on five items, including the Korea AI legislative framework (draft). The session will be livestreamed.[2]
3 November 2026
  • The Personal Information Protection Commission (PIPC) and the Korea CPO Council will hold the 2026 Personal Information Policy Forum on Nov. 9 in Seoul. It will cover the shift to prevention-focused protection under the amended Personal Information Protection Act, stronger CEO/CPO accountability, and autonomous AI security against AI-driven breach threats. [1][2][3]
  • PIPC pre-announced a partial amendment to its organizational enforcement rule (Notice No. 2026-143) on Nov. 3. The notice text does not confirm the contents, including whether an AI privacy team is being created, or the comment deadline; those are only in attached PDFs. [4]
  • LG Uplus signed an agreement with Seoul Women's University on joint privacy and AI research. It will study safe data combination, re-identification risk analysis, and automatic detection and pseudonymization of sensitive information, centered on PIPC's innovation zone at the university. [5]
2 November 2026
  • A bill introduced by Rep. Choi Su-jin and nine others would amend the Industrial Digital Transformation and AI Utilization Promotion Act. It would add a legal basis (Article 12-2) for the industry minister to issue development and use guidelines for industrial AI safety and reliability, citing weak defenses against hacking, improper use of training data, and missing controls on autonomy. The bill is in a public comment period (Nov 4–13) and has not yet been taken up by the committee.[1]
  • The extension of AI-generated content labeling duties to the distribution stage is stalled by a disagreement between MSIT and the media and communications regulator. The regulator wants platform and poster duties added through the Information and Communications Network Act, while MSIT wants generation and distribution handled consistently. The National Assembly subcommittee asked for a revised draft by early October, and none had been submitted by the end of the month.[2]
  • Reports of AI-lowered barriers to hacking are prompting more firms to seek legal and security help. The Personal Information Protection Commission received 579 breach reports in January–August, already above last year's full-year 447. Companies must report within 72 hours, and firms such as Sejong and Kwang Jang (광장) have set up roughly 60-person dedicated teams.[3]
1 November 2026
  • LH obtained NIA's AI ethics compliance mark, which requires an institution to set its own AI ethics principles and have staff complete ethics training. 85.7% of LH's target employees completed the training under guidelines issued in April. [1][2][3]
  • MSIT appointed former Kakao VP Kim Se-woong to head the National AI Service Innovation Task Force. The 61-person unit uses private developers hired as fixed-term civil servants instead of outside contractors. [4]
October 2026 (24 days)
30 October 2026
  • MSIT ran an emergency survey of Korean software and AI firms on how China's proposed AI export restrictions would affect them, asking about adoption of Chinese models (DeepSeek, Alibaba Qwen, Tencent Hunyuan, Moonshot Kimi) and about on-premises and API use. [1]
  • China opened consultation on revising its export ban/restriction technology list on Oct. 16, covering models above 2 trillion parameters and AI agents. Korean industry worries that many local firms fine-tune foreign models, so tighter rules could disrupt development. [1]
29 October 2026
  • MOIS and the Personal Information Protection Commission will hold the third 'Everyone's Discussion' forum on AI-era personal data on Nov. 1, with 200 citizens deliberating. Topics include shifting from prior consent to a 'prior notice, opt-out anytime' model, and results from an online discussion of about 1,200 people will be shared. [1][2]
28 October 2026
  • The Presidential National Science and Technology Advisory Council, via MSIT, has launched a study to build a science-based policing system and national strategy against AI-enabled crimes such as deepfakes and voice phishing, with the study and a cross-ministry strategy targeted for completion by February 2027.[1]
  • The Education Ministry is considering criminal complaints against test-takers who use AI glasses to cheat on next month's CSAT, with obstruction of official duties by deceit (up to 5 years in prison) under discussion. AI glasses are already banned from exam rooms, and radio-wave detectors will not be introduced this year.[2]
27 October 2026
  • The National Science and Technology Advisory Council and the National Police Agency commissioned a study, due by February 2027, to build a full-cycle prevention, investigation and victim-protection system for AI-enabled crime. Police still keep no separate AI-crime statistics, and deepfake sexual exploitation material was about 35% of cyber sexual violence cases last year, roughly 1,500 cases. [1]
  • The Education Ministry met teacher unions on AI smart-glasses cheating ahead of the Nov. 19 CSAT. Unions asked for extra proctors and pre-entry detection devices, while the ministry has proctors check by eye; a union survey found 82.3% of teachers say current procedures cannot identify the glasses. [2]
26 October 2026
  • MSIT will review adding labor-management AI used for delivery-platform dispatch and pay to high-impact AI by Q1 next year, and is also weighing an AI Basic Act amendment. A labor ministry survey of 50 delivery workers found they accepted risky driving under algorithmic pressure, scoring 4.54 out of 5.[1]
  • SK Telecom published its first AI Governance Report, which includes lifecycle guidelines for internal AI agents covering development, testing, operation and retirement. Its chief privacy officer oversees an integrated privacy and AI governance system, and the company was fined by the privacy regulator after last year's USIM data breach.[2]
24 October 2026
  • As AI shortens the time needed to exploit vulnerabilities, patch deployment speed has become the core problem. One cited analysis puts the share of internet-exposed services running unsupported PHP 8.1 or older at 73.5%. Deputy PM and MSIT Minister Bae Kyung-hoon told a National Assembly audit that the government will draw up a government-wide cybersecurity plan. [1]
23 October 2026
  • MSIT and KISA issued a 'caution' private-sector cyber alert on Oct 22 after two critical PHP flaws, reportedly found through AI-assisted research, allowed unauthenticated server takeover; the National Security Office held an emergency meeting and KISA sent urgent security guidance to about 28,000 sites. At the Oct 23 audit, KISA said it had not yet determined how many servers still run vulnerable versions. [1][2][3][4]
  • At the National Assembly audit, lawmakers showed that an AI smart-glasses recording light can be covered with a cheap sticker, and attendees could not tell who was filming. MSIT said a guideline study is under way, and police have confirmed 3 illegal-filming cases involving AI glasses this year. [5][6][7]
  • Lawmaker Lee Jin-sook said redacted personal and non-public information could be recovered with AI in 15 of 75 public agencies checked. Minister Bae Kyung-hoon called it serious if true and promised security checks and stronger cross-ministry cooperation. [8]
22 October 2026
  • MSIT issued a 'caution'-level private-sector cyber crisis alert on Oct 22 over a critical PHP flaw that lets attackers run commands on a server without logging in. The flaw was reportedly found with AI by Calif (using OpenAI's GPT-5.6 Sol) and independently by Wiz, and AhnLab estimates about 110,000 vulnerable servers in Korea. [1]
21 October 2026
  • Lawmaker Koo Ja-keun and nine others introduced a bill to rename the act the Industrial AI Transformation Promotion Act. It would add safety and reliability certification for industrial AI products that lack safety standards, plus an industrial data library with graded security measures. The bill is in committee, with a legislative notice period of Oct. 22–Nov. 5, 2026. [1]
20 October 2026
  • KISA began a national R&D project in August to build AI that verifies whether collected vulnerability reports are real security issues and proposes patch-code candidates. Initial vulnerability detection is out of scope and humans make the final check at each stage. Development is due to finish in the second half of next year, followed by a pilot in KISA's vulnerability-response work.[1]
19 October 2026
  • The Personal Information Protection Commission issued guidance on secondary harm from AI-enabled hacking. It advises individuals to use 2-step verification and check for leaks, and tells companies and agencies to delete abandoned systems and accounts, close unneeded external ports, retire unused APIs, and set up breach-response procedures. [1]
  • NIA designated Korea Airports Corporation and the Korea Employment Information Service as "AI ethics compliance institutions" for adopting AI ethics charters or principles and running staff ethics training. Both plan to add ethics checks or guidelines for future AI services, and the designation comes with a certification mark. [2][3]
  • MSIT and the World Bank co-hosted the Global AI and Digital Summit in Seoul (Oct 19–22), the first outside the US, with ministerial delegations from about 50 countries. The Oct 19 ministerial meeting issued a chair's statement, and Deputy Prime Minister Bae Kyung-hoon pledged international cooperation on responsible AI use. [4]
18 October 2026
  • China released a draft adding 40 core technologies, 23 of them AI, to its export ban and restriction lists, with comments open until Nov 15. A Korean report notes the top Independent AI Foundation Model project (Dokpamo) model scored 47, against 61–63 for leading US models, which leaves Korea exposed to restrictions from both sides. [1]
15 October 2026
  • At a National Assembly audit, MSIT official Choi Woo-hyuk said AI hacking means automated, simultaneous short-burst attacks, and traces of the LLM-based autonomous pentest tool ARTEX were found in the recent financial-sector attack. KISA said it created an AI security division and a vulnerability-management unit, and its AI incident-response system is meant to cut post-incident investigation time to a third or less.[1][2]
  • Raon Secure was selected to lead an MSIT/NIPA-backed physical-AI multilayer security project. It will extend DID identity to robot agents, with permission and delegation tracking so out-of-scope commands can be blocked at execution.[3]
  • Examinees wearing AI smart glasses were caught at Yonsei University and University of Seoul essay exams. KCA is developing a radio-based detector, but it is reportedly unlikely to be ready for this year's CSAT.[4]
14 October 2026
  • MSIT said the Cabinet on the 13th passed amendments to the AI Basic Act and the Data Industry Act. Universities and SMEs get easier access to government-funded AI training data, and the government gets a legal basis to request information on where public and private data are held. [1]
  • Both amendments take effect from April 2027, and MSIT has begun revising subordinate regulations. [1]
13 October 2026
  • PIPC Chair Song Kyung-hee told the National Assembly audit that, after recent breaches, the agency will shift to stronger sanctions and preventive protection. Punitive fines of up to 10% of revenue have applied since September, with new enforcement fines for non-cooperation and evidence destruction, and she urged quick passage of Personal Information Protection Act amendments. [1]
  • MSIT said the cabinet passed amendments to the AI Basic Act and the Data Industry Act, taking effect in April 2027. The AI Basic Act amendment requires government-funded AI training data to be offered through AI Hub, with exceptions for trade secrets, personal data, copyright and national security. [2]
  • A commentary on a Sept. 21 forum talk says the government answered calls to slow AI development by saying Korea, as a catch-up country, cannot afford to decelerate. It says policy has shifted toward 3 mega-projects: semiconductors, physical AI and AI data centers. [3]
12 October 2026
  • MSIT is drafting an AI Basic Act amendment that extends deepfake labeling duties from AI operators to users such as YouTubers, who would insert their own watermarks. Fines for violations are suspended until at least January 22 next year. [1]
  • KMCC Chair Kim Jong-cheol told the National Assembly audit that the commission will phase in revisions to AI user-protection rules, including AI-generated content labeling and recommendation-algorithm transparency, and protect minors from deepfake sexual content. A council of 10 agencies was formed this year to respond to AI-enabled crime. [2][3]
  • Chorokumbrella and the International AI Ethics Association issued a non-binding child-rights AI charter that requires child-rights impact assessments before developing or significantly changing AI that could seriously affect children. It also bans designs that exploit children's emotional dependence, and has 20 articles. [4]
10 October 2026
  • The Ministry of Health and Welfare and the Korean Neuropsychiatric Association issued two guidelines on generative AI use for mental health (one for all users, one for children, teens and guardians) to prevent overreliance and adverse effects. The general guideline has 21 recommendations, including not entering personal information, not treating AI diagnoses as conclusions, and seeking human or professional help in a crisis; a guideline for high-risk groups is planned for 2027. [1]
8 October 2026
  • SK Telecom and Upstage, rivals in the Independent AI Foundation Model project (Dokpamo), launched a sovereign AI cybersecurity consortium with six security firms and three universities. It applies mutual red-teaming between A.X K and Solar, with Korea University, Soongsil University and Pusan National University as third-party evaluators, and discovered weaknesses are reused as training data. [1]
  • Rep. Lee Ju-hee and 11 others introduced an Information and Communications Network Act amendment (bill 2221842) that would make generative-AI chat services show minors overuse warnings and elapsed-time notices, and provide tools to limit usage method, time and time of day on request. The bill is at the receipt stage and its committee has not yet been assigned. [2]
  • The Ministry of Health and Welfare and the Korean Neuropsychiatric Association released guidelines on using generative AI for mental health, covering privacy checks, not mistaking AI advice for medical diagnosis, and seeking professional help in a crisis. A ministry survey found 86.9% of respondents want a government role, and a tailored guideline for high-risk groups is planned for 2027. [3]
7 October 2026
  • President Lee Jae-myung ordered a shift from reactive to preventive security after a string of financial and public-sector data breaches in which AI use is suspected, and the prime minister called for an overhaul of privacy certification. MSIT minister Bae Kyung-hoon said the information-security budget would rise from about 360 billion to 520 billion won, and that the amended network act (in force Oct 1) allows investigations on signs of hacking alone, without a company report. [1]
  • Financial regulators shared 33 attacker IPs (28 unique) with the financial sector and KISA's C-TAS after the bank breaches, and an AI penetration-testing tool, ARTEX, was suspected of being misused. Shinhan Bank's loan-broker service was bypassed, exposing about 25,000 customers' data; KISA said logs leave no trace of generative AI use, so actual use is hard to confirm. Other breaches reported include Kia union members (about 5,000) and Yoido Full Gospel Church members (about 850,000, suspected). [2][3]
  • MSIT plans to release a national AI safety ecosystem master plan in December. Minister Bae agreed that safety institute staffing and budget should grow, after lawmakers noted safety and trust funding of just 27.4 billion won against 6.149 trillion won for AI performance infrastructure. He also said human control mechanisms such as a kill switch are needed, and that the Independent AI Foundation Model project (Dokpamo) will continue. [4]
6 October 2026
  • Seven Korean financial firms were hit by a suspected AI-driven intrusion tool, ARTEX, with about 67,000 records leaked. PM Han Sung-sook ordered neglected government websites shut down, a review of security governance, and a re-check of privacy certification, and police opened a formal investigation. [1][2][3][4][5][6][7][8][9]
  • MSIT's AI Safety Institute ran prompt-injection tests on autonomous AI agents. No LLM was found that blocks attacks completely, with average defense rates of 53.3–93.9% by model, and every tested model failed against malicious commands planted in stored information. [10]
  • At the National Assembly audit, a lawmaker noted that only 3 of 538 approved AI medical devices were submitted to MSIT for a high-impact AI check. The AI Basic Act does not require firms to submit or register their self-assessments, and all 3 reviewed devices were ruled not high-impact, including a brain-hemorrhage reader. MSIT said it would review reflecting the issue in system improvements. [11][12]
5 October 2026
  • AI-assisted hacking spread from major banks to savings banks and capital firms, with personal data leaked from seven financial firms, about 70,000 records. Police are investigating traces of the open-source autonomous pentest tool ARTEX AI. Meanwhile KISA's new AI cyber-threat vulnerability management center saw its requested 20 extra staff and KRW 17 billion project go unfunded in the government budget proposal, and the National Assembly science committee audit on Oct 6 is expected to take this up. [1][2][3]
  • The Youth Counseling & Welfare Institute gave private firms 14,282 pseudonymized youth crisis-counseling records (suicide, self-harm, sexual violence) for AI training on July 29, with no notice to users and no separate consent. Re-identification risk has been raised because conversation text was included and only identifiers were changed. [4]
  • Of 31 models on the government-wide AI common platform, 19 are foreign-made, including BAAI's Bge-M3, which is on the US Entity List. The government admitted it had no vetting criteria for a model's country or developer of origin. [5]
4 October 2026
  • After signs that AI hacking tools were used in a data leak at a Korean financial firm, security experts called for defensive AI systems and a deputy-PM-level cross-government cybersecurity control tower. Cyber functions are currently split across the presidential national security office, MSIT, the National Intelligence Service and the Personal Information Protection Commission, and the experts also urged executive accountability, including CEOs, for breaches. [1]
2 October 2026
  • MSIT's budget submission to the National Assembly merges the Independent AI Foundation Model project (Dokpamo) into a frontier-AI data program. Funding shifts from a 25 billion won grant this year to a 750 billion won equity investment via the Korea Development Bank next year (total 800 billion won, including 50 billion won for cybersecurity AI training infrastructure). Lawmakers' offices say investment-selection criteria, the government's stake and recovery terms, and budget details remain unclear.[1][2]
  • The Ministry of the Interior and Safety finalized public-sector AI ethics standards on Oct 1, built on 6 core values (publicness, equity, transparency, accountability, safety, privacy) with 68 checklist items and a three-tier self-assessment system. The accountability standard states that humans remain ultimately responsible even when AI supports or makes administrative decisions, and it requires human oversight and remedy mechanisms.[3]
  • At the 5th Chief AI Officer (CAIO) council, the National AI Strategy Committee presented a draft AI legislative framework centered on the AI Basic Act. It covers closing legal gaps in defense and security, providing predictability for businesses, and responding immediately to AI harms; attendance expanded to 42 central agencies. The Personal Information Protection Commission pushed for stronger agency-level privacy management after recent hacking and leak incidents, and ministries were asked to help set standards for AI agents' use of personal data.[4][5]
1 October 2026
  • The National AI Strategy Committee held a seminar on calls to slow AI development and promoted a "speed responsibility" approach: Korea should build frontier capability together with safety, security and control. Experts urged independent model evaluation, advanced-AI risk assessment and safety verification, and international coordination; one speaker said existing rules cannot control frontier AI risks. [1][2][3][4]
  • MOIS and the Personal Information Protection Commission will hold the third "Everyone's Discussion" forum on AI-era personal data protection and use on Nov. 1 with 200 citizens. The government plans to announce its policy direction within the year. [5][6]
  • The Personal Information Protection Commission met pseudonymized-data support centers and Personal Information Innovation Zone operators in Busan to gather field input on AI-era data rules. It plans to pursue institutional changes to fit new data-use methods. [7]
September 2026 (20 days)
30 September 2026
  • MSIT will run a government–private working group from October to draft a national AI safety ecosystem master plan, covering an AI safety management framework, response capacity for advanced high-performance AI risks, and safe everyday AI use. About 25 members include the AI Safety Institute, NIA, KISDI and TTA, and the plan is due by December.[1][2][3][4]
  • The Korea Media and Communications Commission plans AI-generated content labeling duties and platform management duties, including disclosure of how recommendation algorithms work and when they change, and user choice over recommendations. It set out 4 goals and 12 tasks, with legal amendments planned within the year.[5]
  • The Personal Information Protection Commission marked Personal Information Protection Day by pledging to overhaul privacy protection for the AI era and strengthen corporate and institutional accountability and prevention systems.[6]
29 September 2026
  • Eight months into the AI Basic Act, the MSIT-backed support desk logged 461 online inquiries, 212 on transparency duties and 117 on high-impact AI. Experts urged clearer legal or guideline criteria on who must apply watermarks and how, and on what counts as high-impact AI.[1]
  • Korea's Personal Information Protection Commission plans to publish criteria on what information and authority can be entrusted to AI agents within the year. The item appeared in coverage of Meta's Muse agent, which uses a payment path that keeps card details from the agent.[2]
  • The Ministry of the Interior and Safety launched the first statutory AI and data administration officers' council, and plans to distribute AI impact assessment guidelines by December. It also plans to set up a public AI and data association by 2027.[3][4]
28 September 2026
  • The Personal Information Protection Commission hosted a Global AI Privacy Forum in Seoul, bringing together international regulators, GPA leadership, and firms like Microsoft, Meta, and Sony to discuss AI innovation, cross-border cooperation, and child/youth privacy protections. [1][2]
  • MSIT is reportedly reviewing whether to halt the Independent AI Foundation Model (Dokpamo) project and instead create a public-private SPC (30% government stake) to fund a frontier-level model, amid concern the shift could favor large conglomerates over startups; over 2000억원 has already been spent and MSIT says no decision on a third evaluation round or the SPC has been made. [3][4][5]
  • MSIT and IITP launched policy research on regulatory frameworks for 6G/AI network infrastructure, selecting institutions in October for 13 new projects running through 2027년 7월, covering AI-autonomous network design/verification standards, human-intervention criteria, and net-neutrality policy for AI environments. [6]
23 September 2026
  • MSIT and NAVER Cloud began developing a Korean cybersecurity-specialized foundation model, backed by 256 Nvidia B200 GPUs for 10 months, aiming to integrate a 700B-scale model built on ~830TB of security data and shift Korea's information security systems to domestic AI by 2027; KISA separately said a Korea-specific security model is needed since only about half of AI-flagged vulnerabilities prove genuinely valid. [1][2][3]
  • The Personal Information Protection Commission held its AI privacy public-private council to discuss agentic AI risks, with interviews across 15 organizations flagging prompt injection and excessive AI authority as top concerns; it plans to define boundaries for autonomous vs. human-approved tasks, retention/deletion rules for logs and memory, and to release an agentic AI privacy guide by year-end ahead of the amended Personal Information Protection Act taking effect in March 2027. [4][5][6]
  • The Intellectual Property Office co-hosted a strategy forum with the Korean Academy of Engineering on IP rights covering AI models, synthetic data, and workplace know-how, and plans institutional reforms to clarify rights and establish fair compensation systems. [7]
22 September 2026
  • MSIT and KISA piloted GPT-5.5 Cyber scans of national critical-infrastructure websites and software, and experts confirmed about 50–60% of the AI-flagged candidates as valid vulnerabilities. Notification time for companies fell from up to a month to within a week, but KISA did not disclose the total count, the target institutions, or the specific flaws. [1]
  • MSIT discussed development plans and government support with the Naver Cloud consortium, which has begun building a cybersecurity-specialized AI foundation model for critical national facilities and industrial sites. Naver Cloud was selected as the project's implementer on September 3. [2]
21 September 2026
  • MSIT finalized a national R&D AI research ethics guide requiring fact-checking, transparency, and disclosure of AI use, banning submission of evaluation materials to external AI and AI-driven distortion of reviews, with 7 recommended practices for researchers [1].
  • The Korea Economic Research Institute published a report on competition issues in the AI agent era, examining bundling, vertical integration, and self-preferencing risks and proposing policy responses for Korea [2].
  • Prime Minister Han convened the first meeting of the Global AI Hub Promotion Committee, reviewing site-selection plans with a public hearing set for October 2026 ahead of a target 2027 hub opening [3].
20 September 2026
  • MSIT and KISTEP issued an AI research ethics guide for national R&D researchers, requiring disclosure of generative AI use in papers and classifying hidden prompts that manipulate peer review as research misconduct. [1][2]
  • MSIT recommended four major companies' nine labs (LG Electronics, HD Hyundai Robotics, Samsung Medison, Samsung Electronics) for the revived AI alternative military service program, seeking 81 researchers against an annual quota of 120 for 2027-2029. [3][4]
18 September 2026
  • KISA agreed to expand AI-related cybersecurity cooperation with the World Economic Forum's Cybersecurity Center and Singapore's Cyber Security Agency, covering AI security talent, threat information sharing, and critical infrastructure protection. [1]
17 September 2026
  • IITP announced it will fund R&D on AI-based intelligent security and autonomous defense to counter expanding attack surfaces from AI agents and physical AI, covering zero-trust and supply-chain security. [1]
  • Deputy PM Bahk Gyung-hoon said Korea cannot slow its AI development pace but must faster build safety/trust systems for deepfakes, cybersecurity and privacy, amid a debate sparked by rankings showing no Korean model in the global top 25 (best domestic model, Motif Technologies' Motif 3, ranked 31st). [2][3]
  • A commentary argued that before proceeding with the 2500 billion won 'Everyone's AI' project, the government and telecoms must first verify and strengthen zero-trust security, 24-hour monitoring, and breach-response systems given recent telecom data leaks. [4]
16 September 2026
  • KISA is using frontier AI models to scan security products' source code for vulnerabilities, providing analysis reports and remediation guidance, and trained about 50 security staff on AI-based vulnerability response [1].
  • Deputy PM Bae Kyung-hoon rejected US-style AI slowdown calls, saying Korea as a follower nation cannot ease AI development speed, while pledging technical and institutional safeguards for deepfakes, cybersecurity, privacy, and jobs alongside AI Act implementation and the next phase of the Dokpamo project [2][3].
15 September 2026
  • Experts urged Korea to pair its AI investment push with safety infrastructure, noting plans to spend 4.7 trillion won on the Independent AI Foundation Model project and frontier AI next year alongside safety evaluation, risk-behavior detection, and emergency shutdown capabilities. [1]
  • MSIT, the transport ministry, police, and Hyundai Motor/Kia ran a joint public-private cyber-incident drill simulating an AI-driven attack that tampered with vehicle software update files, testing detection, isolation, and threat-sharing procedures. [2]
  • Korea's National Data Agency is hosting an OECD/ADB workshop on AI-friendly global data standards in Seoul, drawing data experts from 25 countries to discuss AI-linked data management and a closing hackathon. [3]
14 September 2026
  • Korea's broadcasting and communications regulator is drafting a dedicated AI User Protection Act that would require AI service providers to disclose information and honor user rights such as access, refusal of use, and explanation requests, alongside a planned dispute mediation body. [1]
10 September 2026
  • National Sports Promotion Corporation earned NIA's 'AI Ethics Compliance Institution' certification mark, with over 70% of staff completing AI ethics training and new internal guidelines on public interest, safety, and accountability. [1]
  • SPRi published a study on growth paths for top-tier AI talent, defining core competencies for researchers versus engineers based on interviews and surveys, and recommending expanded K-12 entry pathways and stronger university research infrastructure. [2]
  • SPRi proposed a strategy for cultivating 'entrepreneurial AI talent', based on a survey of 715 advanced AI professionals and startup founder interviews, recommending graduate-level entrepreneurship training and startup-friendly ecosystem support. [3]
9 September 2026
  • The Financial Supervisory Service held a CISO meeting across all financial sectors on frontier AI-enabled cyber threats, urging fast action on basic security such as asset management and immediate patching. It plans AI usage guidelines, stronger executive accountability and on-site inspections, with strict action against weak institutions. [1]
  • The National Assembly Research Service warned that children and teens face risks from AI chatbots, including emotional dependence and overuse that can lead to high-risk behavior. It called for legislation and policy that reflect these risks, citing overseas moves on safety-by-design and operator responsibility. [2]
  • The National AI Strategic Committee's anniversary reporting says it will pursue improvements to the Independent AI Foundation Model (Dokpamo) evaluation system and a K-AI cybersecurity framework. It also plans GPU allocation principles and a stronger cross-ministry coordination role. [3][4]
8 September 2026
  • A promulgated amendment to the Personal Information Protection Act adds an AI-development exception allowing personal data that is hard to anonymize or pseudonymize to be used beyond its original purpose. Use requires safeguards and a Personal Information Protection Commission (PIPC) review, with risk assessment and conditions where needed; the law takes effect March 9, 2027. [1]
  • MSIT began a study on safe-use guidelines for AI glasses, covering exam cheating and unauthorized filming. The PIPC has also started drafting personal-data processing principles for AI glasses. [2]
  • The government will split 170 billion won in Everyone's AI GPU support among the SK Telecom, KT and Kakao consortia according to user numbers and service usage. The first evaluation is slated for as early as March 2027, and low scorers will not be dropped. [3]
7 September 2026
  • MSIT and IITP are budgeting about 3 billion won to set up a joint AI research lab in Europe next year, tasked partly with ethics/safety verification research and helping Korean firms respond to EU AI Act requirements, linking it to the existing New York AI Frontier Lab. [1]
  • At a defense AX and security conference, experts called for integrated governance across the Ministry of National Defense, MSIT, and the National Intelligence Service to build AI-based defense systems beyond K-RMF and zero-trust as drones, autonomous weapons, and cyber warfare converge. [2]
4 September 2026
  • MSIT selected a Naver Cloud-led consortium as the security-specialized track of the Independent AI Foundation Model project (Dokpamo), with 33 industry-academia-public institutions building two 700B-class MoE models for cyber offense/defense. [1]
  • Korea's AI policy command structure was reshuffled with new presidential and national AI committee appointments, and experts are pushing to redesign the Dokpamo project and link it to a frontier-model initiative. [2]
  • The same reshuffle also puts under review whether AI Framework Act regulations will take effect after next year's grace period ends, alongside complementary measures. [2]
3 September 2026
  • MSIT selected the Naver Cloud consortium to build a cybersecurity-specialized AI foundation model, backed by government-supplied GPUs with a mid-term evaluation after 5 months deciding continued support. [1][2][3][4][5]
  • The Financial Services Commission finalized a second emergency easing of network-separation rules, letting 75 financial firms test AI and security SaaS tools to detect vulnerabilities, with results to feed into future AI security guidelines. [6]
  • The Prime Minister's Office held its second field meeting on AI administrative innovation, where officials proposed expanding immunity for AI use and revising incentive/responsibility structures for civil servants. [7]
2 September 2026
  • Korea's Fair Trade Commission issued a revised notice requiring prior substantiation for AI and new-tech product performance claims, effective September 3, and cut the evidence-submission extension window from 30 days to 15 days, with ad-suspension orders for non-compliance. [1]
  • The National AI Strategy Committee met with a European high-level labor-policy delegation to discuss a socially legitimate national AI strategy, tying the government's 'Everyone's AI' push to employment, inequality, and public-trust concerns. [2]
1 September 2026
  • KISA and MSIT will pilot an AI penetration-testing agent in the second-half national cyber crisis response drill, simulating website intrusion, phishing, and DDoS scenarios for consenting firms from October 19–30, with sign-ups open until October 2. [1][2]
  • Lawmaker An Sang-hun and 10 others proposed an amendment to the AI Basic Act requiring conversational AI operators to clearly notify users they are interacting with an AI, with violations subject to fact-finding and corrective orders; the bill is in legislative notice through September 12. [3]
  • The Korea Capital Market Institute reported that eased network-separation rules have expanded generative AI use in the financial sector into investment, credit assessment, and customer service, urging regulators to track performance and feed accumulated risk experience into future rulemaking. [4]
August 2026 (23 days)
31 August 2026
  • KISA and KCA, with a National Intelligence Service branch, held the 'AI Privacy Challenge', in which regional university students used generative AI to find vulnerabilities in personal-data processing systems while agency security staff practiced threat detection, blocking and recovery; 3 teams won awards. [1][2]
  • The government named operators for Everyone's AI (the national public chatbot) on Aug. 28, a service meant to act as an AI agent handling tasks such as applying for subsidies and issuing 83 types of public certificates. A Chosun Ilbo column warns that users must hand over more personal data and permissions, and argues that storage and use of that data should be disclosed transparently. [3]
30 August 2026
  • Korea's presidential office restructured its AI policy leadership, naming Lee Hae-min as AI future planning secretary, reappointing Ha Jung-woo as full-time vice chair of the National AI Strategy Committee, and confirming Bae Kyung-hoon as science minister overseeing the Independent AI Foundation Model project and national AI computing center. [1][2][3][4]
29 August 2026
  • MSIT selected SK Telecom, KT, and Kakao as operators for the 'Everyone's AI' free public AI service, with each firm developing GPU-supported search, booking, and payment features to launch later this year. [1]
28 August 2026
  • MSIT's Independent AI Foundation Model project (Dokpamo) second-round evaluation eliminated Motif Technologies despite its top benchmark score, leaving SK Telecom, Upstage and LG AI Research; Motif filed an objection demanding detailed scores and a re-review. Total scores were SK Telecom 70.6, Upstage 69.9, LG AI Research 69.0 and Motif 65.8; MSIT says it will decide within 15 days, and criticism over scoring and transparency continues, with the minister saying he will consider improving the evaluation system.[1][2][3][4][5]
  • MSIT and NIA released 29 datasets from the five Independent AI Foundation Model teams, including AI safety data, free of charge via AI Hub.[2]
  • The Personal Information Protection Commission held its 4th mock-trial contest on AI and privacy, where teams argued whether AI-generated false sensitive information counts as personal information, along with agentic AI accountability and cross-border transfers. The event drew 54 teams.[6][7]
27 August 2026
  • Korea's AI service user-protection public-private council held its second meeting to review illegal/harmful content risks and regulatory gaps, with Kakao presenting its Kanana Safeguard guardrail model for safety verification. [1][2]
  • The International AI Ethics Association urged Korea's National Education Commission to fully reconsider using AI for first-pass grading of essay-type exam answers, citing bias, lack of explainability, hallucination risk, and inadequate remedy procedures. [3]
  • MSIT and NIA released 29 AI training datasets (about 35.44 million items, ~1.56 trillion tokens) built under the Independent AI Foundation Model project by five selected teams, after quality and safety verification, for free public use via AI Hub. [4][5]
26 August 2026
  • Korea's Independent AI Foundation Model project advanced three teams — SK Telecom, LG AI Research, and Upstage — past its second evaluation round based on benchmark, expert, and user scoring, while Motif Technologies was eliminated. [1]
  • ETRI joined the Agentic AI Foundation under the Linux Foundation to help shape agentic AI standards such as MCP, working alongside roughly 250 member companies on accuracy, security, and governance workstreams. [2]
  • Korea's National Assembly Research Service flagged implementation gaps in the newly effective AI and Data-Based Administration Act, citing unresolved issues in data quality, AI reliability/accountability, and governance/security ahead of its August 28 enforcement date. [3]
25 August 2026
  • AhnLab warned that "memory poisoning" attacks can corrupt an AI agent's long-term memory to distort future decisions, and Korea's privacy watchdog said it will draft AI agent guidelines and related legislation in the second half of 2026. [1][2]
  • The government finalized Korea's AI ethics principles, built on three values and seven principles (human-centeredness, privacy, fairness, accountability, safety, reliability, transparency), explicitly naming users as responsible actors alongside developers, though the code remains a non-binding voluntary norm under the AI Basic Act. [3][4][5]
  • ETRI joined the Linux Foundation's Agentic AI Foundation (AAIF), a roughly 250-member group including Google and OpenAI, to help set international standards for agentic AI protocols such as MCP. [6]
24 August 2026
  • MSIT finalized Korea's new AI ethics principles, built on three values and seven principles, explicitly designating users, government, and civil society as "AI actors" alongside suppliers, and requiring human control over AI agents plus stronger accountability for large AI providers. [1][2][3][4][5][6][7]
  • Commentary flagged that because the revised principles remain a non-binding self-regulatory code, expanded civic oversight and eventual legislation are needed to make them effective against harms like job loss and privacy violations. [7][1][2][3][4][5]
22 August 2026
  • Korea amended its Personal Information Protection Act to allow use of personal data for AI development when pseudonymized or anonymized data is insufficient, subject to review by the Personal Information Protection Commission and mandatory prior risk assessment for sensitive data; the change takes effect 6 months after promulgation. [1]
21 August 2026
  • South Korea's National Assembly passed an amendment to the Personal Information Protection Act allowing lawful use of personal data for AI development when pseudonymized or anonymized data is insufficient, subject to enhanced safeguards and Personal Information Protection Commission review; use of sensitive or unique identifier data requires prior risk assessment and disclosure, with the law taking effect 6 months after promulgation. [1]
20 August 2026
  • South Korea's National Assembly passed an amendment to the Personal Information Protection Act allowing original personal data to be used for AI training when public interest is recognized, subject to review by the Personal Information Protection Commission, mandatory prior risk assessments for high-risk data processing, and public disclosure requirements. [1]
19 August 2026
  • MSIT's second-round evaluation of the Independent AI Foundation Model project (Dokpamo) passed Upstage, SK Telecom, and LG AI Research to the next stage while eliminating Motif Technologies, based on benchmark, expert, and user-evaluation scores. [1][2][3][4][5]
  • Critics noted the government did not disclose team-by-team scores or rankings from the tax-funded Dokpamo evaluation, despite using a 40/35/25-point benchmark-expert-user rubric. [6]
  • MSIT is reviewing a broader restructuring of the Dokpamo and related AI programs, including future competition rules, resource allocation, and whether a fourth selection stage will proceed. [1]
18 August 2026
  • South Korea's Independent AI Foundation Model (Dokpamo) project eliminated Motif Technologies in its second-stage evaluation despite top benchmark scores, advancing Upstage, SK Telecom, and LG AI Research to the third round; the government said a review with global evaluator Artificial Analysis found no evidence of benchmark-targeted overfitting. [1][2][3]
  • The Korea Institute of Finance called for financial firms to establish AI governance frameworks covering fairness, transparency, consumer protection, and data security, urging regulators to set minimum explainability standards for black-box models while letting firms implement governance autonomously via sandboxes. [4]
17 August 2026
  • MSIT's revised second draft of AI ethics principles reportedly removed or weakened anti-discrimination provisions covering gender, age, and disability, as well as developer/provider accountability language present in the first draft. The Korean Women's Associations United criticized the draft for shifting bias/discrimination responsibility onto users and demanded explicit non-discrimination and inclusion principles with clear duties for each actor across the AI lifecycle. [1]
14 August 2026
  • MSIT unveiled at an Aug 14 forum a draft revision of the 2020 AI ethics standards, setting out three core values and 7 principles including human-centeredness, privacy protection, fairness, safety and reliability. The draft is a non-binding soft norm, and MSIT, working with KISDI, reportedly plans to publish the final principles this month. [1][2][3]
13 August 2026
  • South Korea's government proposed an amendment to the Personal Information Protection Act that would expand permissible use of personal data for AI development when enhanced safeguards are in place, subject to review by the Personal Information Protection Commission; the bill is currently under National Assembly deliberation after passing preliminary readings in May and July 2026. [1]
12 August 2026
  • The Board of Audit and Inspection found duplicated and poorly quality-controlled public AI training datasets built with more than KRW 1.6 trillion in government funds. It recommended that MSIT and other agencies review duplication, share plans in advance, and adopt unified quality standards. [1]
11 August 2026
  • KISA and IITP convened an industry council to shape information-security R&D priorities, focusing on AI agent security, zero-trust, and physical-AI risks. [1]
  • The Korea Institute of Finance issued a brief urging regulators to prepare consumer-protection rules for AI shopping agents before they reach retail financial customers, citing disclosure and conflict-of-interest risks. [2]
9 August 2026
  • MSIT announced an 'Agentic AI Initiative' covering physical AI safety, including plans for safety guidelines and sandbox support for humanoid robots after a forum raised concerns about malfunction risks from physical manipulation. [1]
7 August 2026
  • Korea lacks pre-release safety verification for AI models: MSIT's AI Safety Institute evaluated 42 domestic and foreign AI models but found no cases of pre-launch review, unlike the US system where firms submit models to government for pre-release assessment. [1]
6 August 2026
  • MSIT and KISA published an AI security red-teaming guide to help companies conduct pre-emptive risk checks on training data, core models, applications, and autonomous agent behavior. [1]
3 August 2026
  • Korea's Personal Information Protection Commission urged Samsung Electronics and Meta to build in privacy safeguards for upcoming AI glasses, such as recording-indicator alerts to protect bystanders, and plans to establish related privacy principles this year. [1]
2 August 2026
  • AI-Ethics-KR criticized Korea's autonomous vehicle law and the proposed physical AI special act for including vague "unavoidable" personal-data-use exemptions that prioritize industry competitiveness over clear standards or civil-society input. [1]