알림 목록

Ulsan City has set a goal to increase the AI-based digital transformation rate of local companies from approximately 7% to 40% by 2030, and launched the AI ​​Capital Promotion Headquarters as a separate organization in 2026. Although AI has expanded from partial informatization projects to a comprehensive urban strategy encompassing industry, transportation, administration, and investment attraction, the City Council’s verification unit remains limited to departmental work reports and individual budgets and ordinances. An integrated verification system connecting declarations to results has not been established, and by 2028–2029, it will become difficult to reconstruct the discrepancy between the number of executed projects and the actual corporate transformation rate. Consequently, the council's verification structure for Ulsan's AI city strategy is lagging behind the speed of its implementation.

In 2026, the Industry and Construction Committee reviewed the AI ​​Capital Promotion Headquarters' data center, autonomous buses, smart city services, the Sovereign AI Cluster, and promotional projects in a single meeting. Although the scope of verification expanded simultaneously to include infrastructure, industrial transformation, citizen services, and city branding, the verification results were fragmented across schedule, differentiation, utilization rates, satisfaction, and promotional effectiveness. No common performance axis was identified to assess the sequential relationships between projects or cumulative financial risks, and there is a growing risk that the normal execution of individual projects over the next two to three years will be tallied as the overall performance of the city strategy. While current legislative activities are reviewing AI projects, they have not yet reached the stage of verifying the AI ​​city as a single performance portfolio.

During the 261st Industry and Construction Committee meeting, questions were raised regarding the power-saving competitiveness of underwater data centers, the commercialization schedule for autonomous buses, the utilization rates and satisfaction of smart city services, and the distinctiveness of the Sovereign AI Cluster during the business report of the AI ​​Capital Promotion Headquarters. While the Council’s interest has shifted from verifying project names and budgets to demonstrating effects and usage indicators, there is no evidence that the results of these inquiries have been accumulated into a Council Performance Board that includes baseline values, target values, and verification timelines. If the same project undergoes organizational restructuring or a name change by 2028, the continuity between previous inquiries and subsequent outcomes weakens. Although an awareness of the need for performance verification has emerged, the level of institutionalization remains in its early stages.

During the 263rd Industrial Construction Committee meeting, while reviewing the ordinance bill to eliminate blind spots in route bus right turns using AI, the committee examined product pricing, distribution status, discrepancies in budget estimates, and the technical effectiveness of the pilot project. Although the review process prioritized examining price appropriateness and demonstration results over the designation of "AI," the publicly available data does not confirm termination conditions for re-examining demonstration results or subsequent reporting cycles following the original bill's approval. If the pilot project is expanded over the next two to three years, the causal relationship between the initial verification questions and the expanded budget could be severed. While prior control over individual technologies is confirmed, the continuity of post-hoc control remains unconfirmed.

In 2025, the City Council reviewed the "Ordinance on the Implementation of AI-Based Administration in Ulsan Metropolitan City" and conducted training on policy analysis using AI and the utilization of legislative data at council members' workshops. Although both the subjects of regulation and the means of verification have simultaneously entered the realm of AI, results showing that the Council's own analytical capabilities have been permanently internalized into budget, audit, and legislative evaluation have not been disclosed. If the gap between the executive branch's technological and data capabilities and the Council's verification capabilities widens around 2028, information asymmetry will become structural. Legal interest and learning have begun, but independent verification capabilities are still in the process of being formed.

At the 261st plenary session, a city goal was presented to raise Ulsan’s corporate AI adoption rate to 40% by 2030, while standing committees during the same session discussed projects related to AI data centers, cloud computing, corporate support, and talent development. Although the city goal was presented as a result indicator, the project evaluation was fragmented, focusing on individual inputs and implementation plans. A contribution structure linking how much each project contributes to the increase in the adoption rate was not identified, making it difficult to attribute the cause to individual projects even if the target is not met in two to three years. The Ulsan City Council acknowledges the AI ​​City Declaration but is failing to extend accountability for performance down to the project level.

There is a 33 percentage point gap between the 7% and 40% corporate AI adoption rates. However, the scope of companies subject to measurement, criteria for recognizing AI adoption, the distinction between simple solution usage and process transformation, and the survey frequency and verification bodies were not presented together in the congressional disclosure data. Even if targets are quantified, if measurement rules are not fixed, performance indicators are used as justification for organizational and budget expansion, and their ex-post judgment power is weakened. If the survey population or definitions change after 2028, time-series comparisons will be compromised. Currently, the 40% target is a political directional indicator and cannot be considered established as an auditable performance indicator.

The Industry and Construction Committee inquired about the promotional effect of the 150 million won production cost for an AI documentary, the price appropriateness of AI-based bus safety devices, and cost estimates totaling 40 billion won, including AI investment sectors. While fiscal controls are in operation to verify the validity and unit prices of individual expenditures, the total fiscal exposure—aggregating promotion, infrastructure, corporate support, and talent development into a single AI city portfolio—is not being verified. If AI-related budgets spread across various bureaus and government-funded institutions over the next two to three years, there is a high likelihood that duplicate projects and indirect costs will be omitted during individual reviews. Currently, budget verification is at the stage of item-by-item control, not the stage of verifying capital allocation for the entire strategy.

During the 261st business report, the possibility of overlapping responsibilities between the Corporate Investment Bureau and the AI ​​Capital Promotion Headquarters was raised; the executive branch responded by stating that each department has a distinct role and that the headquarters coordinates the big picture. Although organizational boundaries were recognized as an issue, no project lists, beneficiary companies, or comparison tables of common KPIs were presented to determine whether overlap had occurred. Over the next two to three years, if AI corporate support is combined with RISE, Technoparks, investment attraction, and support for SMEs, multiple support for the same company and duplicate calculation of performance results will accumulate. The Council identified the risk of overlap but did not proceed to an independent verification beyond the executive branch's explanation.

During the review of the AI-based route bus safety ordinance, discussions emerged regarding reviewing expansion based on the results of the pilot project and verifying its technical effectiveness. While the phased approach of expansion following demonstration has been secured, the passing criteria—such as accident reduction rates, false positive rates, device utilization rates, driver acceptance, and cost per vehicle—are not confirmed in the publicly released meeting results. Pilot projects without established standards allow the completion of installations and the number of operational cases to substitute for actual results, thereby increasing the political and financial costs associated with halting the expansion after two to three years. Currently, while the council's awareness of a phased review process is confirmed, the binding judgment criteria for expansion remain unconfirmed.

Inquiries were raised regarding the commercialization schedule and outsourced operation system for autonomous buses, while usage rates and satisfaction surveys were mentioned for smart city services. Although indicators for the operational phase were discussed, a performance structure combining the number of safety interventions, frequency of manual switching, changes in operating costs, and benefits compared to existing modes of transportation was not presented. As pilot routes and services expand by 2028, there is a growing possibility that citizen perception and technological operational performance will be tallied separately. While the Ulsan City Council’s verification of the demonstration progressed to the questioning stage, a comparable judgment system for termination or expansion was not established.

During the review of the AI-based route bus safety ordinance, discussions emerged regarding reviewing expansion based on the results of the pilot project and verifying its technical effectiveness. While the phased approach of expansion following demonstration has been secured, the passing criteria—such as accident reduction rates, false positive rates, device utilization rates, driver acceptance, and cost per vehicle—are not confirmed in the publicly released meeting results. Pilot projects without established standards allow the completion of installations and the number of operational cases to substitute for actual results, thereby increasing the political and financial costs associated with halting the expansion after two to three years. Currently, while the council's awareness of a phased review process is confirmed, the binding judgment criteria for expansion remain unconfirmed.

Inquiries were raised regarding the commercialization schedule and outsourced operation system for autonomous buses, while usage rates and satisfaction surveys were mentioned for smart city services. Although indicators for the operational phase were discussed, a performance structure combining the number of safety interventions, frequency of manual switching, changes in operating costs, and benefits compared to existing modes of transportation was not presented. As pilot routes and services expand by 2028, there is a growing possibility that citizen perception and technological operational performance will be tallied separately. While the Ulsan City Council’s verification of the demonstration progressed to the questioning stage, a comparable judgment system for termination or expansion was not established.

The Industry and Construction Committee inspected the regional specialized manufacturing data activation project and the advanced materials data analysis infrastructure at Ulsan Technopark, and inquired about the status of the AI ​​cloud support project and the AI ​​dissemination project for SMEs in 2026. Although the scope of the council's inspection shifted from equipment and infrastructure to corporate utilization, changes in productivity, defect rates, energy costs, sales, and employment before and after adoption by supported companies were not linked to the disclosed legislative results. There is a risk that if only the number of beneficiary companies is accumulated over two to three years, simple usage and process redesign will be counted as the same conversion performance. While on-site inspections of Industrial AX exist, verification of additional performance at the company level is lacking.

The fact that performance goals and follow-up management regarding AI talent securing employment at local companies and settling in Ulsan were demanded at the meeting serves as evidence that the final outcomes of talent development have been recognized. Conversely, cohort data distinguishing between trainees who completed the program, those who found employment, those employed in AI roles within Ulsan, and those who remained for one to two years has not been confirmed. Even if the number of trained personnel expands by 2028, if the local retention rate is not separated, the performance of the education program and the results of industrial workforce supply will be mixed. While the City Council pointed out the talent drain gap, verification mechanisms to continuously track this are not yet visible.

The council inquired about the construction schedule of the Ulsan AI Data Center, strategies for attracting related companies, corporate demand, and the potential for power savings from the underwater data center. While the scope of verification has broadened from the infrastructure construction itself to demand and ripple effects, performance criteria distinguishing between local companies' computational access costs, usage, utilization rates, and the proportion of external sales were not confirmed. If site, electricity, and construction costs become fixed within two to three years, the costs of correcting insufficient demand or deficiencies in local accessibility will skyrocket. Although the council acknowledged the infrastructure risks, it was unable to finalize a judgment line linking construction approval with regional AX performance.

During the evaluation of underwater data centers, preliminary analyses regarding seawater temperature, ecosystems, and the fishing environment for fishermen were also raised. Although technological competitiveness and regional externalities were included as subjects of evaluation, a baseline for simultaneously comparing environmental, energy, and industrial performance was not disclosed. If environmental impact and industrial performance are managed by different committees or departments after the facility becomes operational, the responsibility for overall profit and loss becomes fragmented. By 2028–2029, sunk costs will act as pressure to prioritize business continuity over environmental and demand risks. While concerns regarding large-scale AI infrastructure exist, the density of verification prior to irreversible investment is insufficient.

The 2025 council members' workshop covered policy analysis using AI and the utilization of legislative data, and the City Council discloses meeting minutes, committee results, and administrative audit action results online. While the foundation of analysis tools and public data exists, a council dashboard that automatically links budgets, contracts, KPIs, audit findings, and action results for AI projects has not been found. As agenda items and documents accumulate over two to three years, it becomes difficult to detect duplication and repetitive issues between projects relying solely on manual searches by council members and policy support officers. Although the Ulsan City Council has begun AI learning, it is difficult to conclude that it has expanded into a permanent verification system.

While the executive branch operates a dedicated AI organization and numerous specialized projects, independent experts or permanent verification units specialized in AI, data, and algorithm audits are not identified in the legislature's publicly available data. In a structure where the technical information asymmetry between the executive and legislature is widening, short-term workshops cannot guarantee the continuity and reproducibility of expertise. If model performance, data rights, cloud costs, and vendor lock-in become settlement issues around 2028, identifying the causes will be delayed relying solely on existing financial reviews. The legislature's own AX is in the early stages of tool adoption, and structural changes to its oversight capabilities are incomplete.

Simultaneously, the proposal of ordinances, work reports from the AI ​​Capital Promotion Headquarters, inquiries regarding data centers, autonomous driving, and smart cities, and demands for indicators of corporate demand and talent retention are confirmed. While the scope of the Council's perception of AI has broadened from a single technology project to an issue encompassing industry, administration, transportation, and talent, a common verification system linking declaration, budget, execution, results, and follow-up measures has not been identified. If the number of projects increases over the next two to three years, it will become difficult to separate cumulative achievements from failures using the current method of questioning by session. The level of preparation is assessed as upper-middle in problem awareness and lower-middle in institutionalization .

Industry, transportation, and smart cities were handled by the Industry and Construction Committee; the implementation of AI administration by the Administration and Local Government Committee; and generative AI education by the Education Committee. While the AI ​​agenda has spread across multiple standing committees, evidence of joint verification of AI performance across the entire city or the integration of indicators between committees has not been confirmed. If interdependence on data, talent, and infrastructure among projects intensifies by 2028, the boundaries of standing committees could become entrenched as a dividing line in performance accountability. The expansion of the agenda is confirmed, but the expansion of integrated oversight remains unconfirmed.

Results-oriented questions regarding utilization rates, satisfaction, the effectiveness of pilot projects, price appropriateness, the retention of local talent, and corporate demand were raised in various meetings. However, there is no continuous record of these questions being reconfirmed using the same indicators in subsequent budget, settlement, or administrative audits. When one-time inquiries are repeated for two to three years, the number of criticisms increases, but performance standards do not accumulate. While performance verification has spread to individual questions from council members, it has not become established as a repeatable procedure within the institution.

The City Council has secured a documentary basis for verification by disclosing meeting minutes, committee results, legislative proposals, and administrative audit data. However, there is no evidence publicly available in a single data structure that covers total AI-related budgets, project-specific KPIs, beneficiary overlaps, contract changes, model performance, and follow-up measures. If project names and responsible departments change after 2028, the cost of restoring the lineage of past performance and current expenditures will skyrocket. While public documents exist, datasets for performance audits are currently absent.

Between the launch of the AI ​​Capital Promotion Headquarters in 2026 and the target of a 40% corporate AI adoption rate by 2030, the period of 2027–2028 marks the first time that interim results in infrastructure, corporate support, and talent development are accumulated. If baselines and contribution measurements are not stabilized during this period, only the target achievement rate will remain by 2029, and the effectiveness of individual projects will remain unidentifiable. Delays in the verification system translate into not merely administrative delays, but the loss of performance history. Currently, the remaining timeframe for the "golden time" is estimated to be approximately 24 months .

For data centers and industrial complexes, site, power, and construction contracts are fixed, while corporate support and talent development accumulate through a multi-year implementing agency and subsidy structure. If investment expands under weak prior performance criteria and conditions for mid-term termination, sunk costs and stakeholders increase simultaneously. Even if a lack of performance is confirmed in 2028–2029, it is highly likely that project downsizing will escalate into issues related to local politics, employment, and contracts. The irreversibility of Ulsan’s AI City strategy is assessed as high risk for infrastructure, medium-to-medium risk for support projects, and high risk for the verification system .

The Ulsan City Council did not stand idly by regarding the declaration of an AI city. It expanded the scope of verification by questioning organizational overlap, technology pricing, demonstration effects, corporate demand, talent retention, environmental impact, and usage indicators.

However, current verification remains limited to the units of sessions, standing committees, and individual projects. A continuous verification line running through urban goals, total budget, project contributions, corporate performance, and follow-up measures has not been established.

The final judgment is “Active AI Oversight + Portfolio Accountability Gap .” While the Ulsan City Council is reviewing AI projects, it is too early to conclude that it is independently and cumulatively verifying the actual performance of the AI ​​city.

Overall Score: 48 / 100

Golden Time Status: ORANGE–RED
Time Window: Approx. 24 months
Failure Mode: Verification illusion where normal project execution replaces city-wide performance
Irreversibility: Rising from medium-risk to high-risk

Evidence Sources

• Ulsan Metropolitan City Council 261st Industry and Construction Committee Meeting—AI Capital Promotion Headquarters Work Report
• Minutes of the 261st Industrial and Construction Committee Meeting of the Ulsan Metropolitan City Council—Business Investment Bureau Work Report
• Minutes of the 2nd Meeting of the 261st Plenary Session of the Ulsan Metropolitan City Council
• Ulsan Metropolitan City Council 263rd Industry and Construction Committee Session—Review of AI-based Bus Safety Ordinance and Supplementary Budget
• Minutes of the 2nd Meeting of the Industry and Construction Committee, 263rd Session of the Ulsan Metropolitan City Council
• Ulsan Metropolitan City Council Industry and Construction Committee on Ulsan Technopark Site Inspection
• 2025 Ulsan Metropolitan City Council Members' Workshop—Utilization of AI Policy Analysis and Legislative Data
• 252nd Industrial Construction Committee Meeting—Criticism of Ulsan AI Status and Company Outcomes

Structural Insight — Congress’s annual accounting hours conflict with AX’s cumulative hours

Local councils verify the executive branch through an annual accounting cycle consisting of budget proposals, supplementary budgets, settlements of accounts, and administrative audits. On the other hand, AX performance manifests over multiple years through data accumulation, process learning, organizational redesign, and enterprise expansion. Even for the same project, the evaluation target shifts from the implementation rate in the first year to the usage rate in the next, and then to productivity thereafter.

This time-axis discrepancy creates a structure where cumulative failures are missed, even if the council conducts thorough reviews every year. Even if the budget execution rate in the year of establishment and the utilization rate in the year of operation are each judged to be normal, the entire project is considered a failure if final productivity or regional residual value is low. The sum of annual normal judgments does not guarantee strategic success.

Therefore, the supervisory risk of Ulsan AI City is amplified by time-based discrepancies rather than a lack of interest from council members. If the disconnect between annual accounting reviews and AX performance over three to five years persists, failures accumulate without clearly manifesting in any single fiscal year. The greatest structural risk of the Ulsan City Council lies not in a failure to verify, but in its inability to determine overall failures despite verifying every year.