The Committee appreciates and supports the government’s “Information Security is National Security 2.0” strategy, which emphasizes the importance of transitioning to Zero Trust Architecture (ZTA) Architecture to strengthen verification protocols and enhance digital protection. However, the reduction in the government’s security budget for 2025 poses a challenge to fully implementing this architecture. Adequate funding is essential for effective identity verification mechanisms, device authentication, and trust inference.
We suggest the government allocate additional resources or adjust existing budgets to strengthen information security protections. In alignment with the national goal of transforming Taiwan into an AI island under the AI Taiwan Action Plan 2.0, we recommend adopting a principles-based framework that balances AI innovation with risk management, aligns with international standards, and emphasizes public-private collaboration.
The Committee also recommends launching cross-agency initiatives to enhance Taiwan’s position in the global quantum computing landscape. Quantum computing could transform key industries and significantly boost competitiveness. We propose a new five-year plan focused on strengthening R&D, cultivating talent, incentivizing industry participation, and developing post-quantum cryptography strategies.
The Committee encourages continued dialogue and collaboration to strengthen Taiwan’s technological resilience and security.
Suggestion 1: Modernize Taiwan’s privacy laws to balance economic development and personal data protection.
The Committee commends the Preparatory Office of the Personal Data Protection Commission (PDPC) for its ongoing efforts to establish the PDPC by August this year. As part of this process, the Preparatory Office is also proposing amendments to the Personal Data Protection Act (PDPA). We recommend leveraging this opportunity to modernize the legislation to reflect Taiwan’s evolving economy and reliance on international trade. We offer the following recommendations:
- Clarify the distinction between data controllers and data processors. Article 4 of the PDPA currently does not differentiate between the roles and responsibilities of data controllers and data processors. To align with international standards and clarify legal responsibilities, we recommend defining a data controller as a person who determines the purposes and means of processing personal data, and a data processor as one that processes data on behalf of the controller.
Processors should be responsible only for implementing appropriate security measures and complying with the terms stipulated in agreement with the controller. Because they typically lack direct relationships with data subjects and insight into the broader context, it is impractical to assign them the same legal responsibilities as controllers.
Clarifying these roles will strengthen accountability, improve compliance, and align Taiwan’s data protection framework with internationally recognized practices. Although this issue was raised in the 2024 White Paper, no substantial progress has yet been made. - Assign breach notification obligations solely to data controllers. Building on the recommended role distinctions in Suggestion 1.1, the PDPA should clarify that data controllers and not data processors are responsible for notifying the competent authority and affected data subjects. Processors should be required only to report breaches to controllers. These obligations should take effect only after the relevant subsidiary legislation outlining thresholds, mechanisms, and timelines is finalized, so as to prevent ambiguity in compliance expectations.
- Make data subject breach notification contingent on a harm threshold. An effective data breach notification mechanism should alert data subjects to actual risks and provide the information needed to mitigate potential harm. In line with the EU’s General Data Protection Regulation and Singapore’s Personal Data Protection Act, notification obligations should be subject to a harm threshold. This prevents over-notification, which may desensitize recipients by alerting them to incidents that pose no genuine threat, ultimately diminishing the impact of critical warnings.
- Grant a rectification period before imposing penalties for violations of data breach reporting and record-keeping requirements. The latest PDPA amendments impose penalties for certain violations, such as failures in data breach reporting and record-keeping, without first allowing a period for correction. However, these violations are rarely intentional, and organizations typically need time to assess the nature and impact of a breach before determining their compliance obligations. Emerging digital businesses may also face uncertainty about which regulatory authority has jurisdiction, leading to inadvertent delays. In line with the approach taken for notifying data subjects of breaches, a rectification period should likewise apply to reporting and record-keeping violations.
- Adopt a flexible approach to security maintenance mechanisms. When establishing security maintenance requirements and management mechanisms, the government should avoid overly prescriptive requirements and allow non-governmental agencies to adopt security measures flexibly based on their industry characteristics, risk exposure, and the types of data they process. Subsidiary legislation should support flexible, risk-based implementation tailored to each organization’s context.
- Establish a transition period of at least one year to support compliance with new PDPA requirements. Non-governmental agencies, particularly those with complex operations, will require sufficient time to implement the systemic changes needed to comply with new data protection requirements, especially those related to data breaches. To ensure effective compliance, we recommend establishing a transition period of at least one year following the promulgation of any new requirements.
- Conduct thorough public consultations on proposed PDPA subsidiary legislation. The latest PDPA amendments grant the competent authority broad regulatory powers to develop detailed implementation requirements. To ensure that all relevant stakeholders have a meaningful opportunity to provide input, the Personal Data Protection Commission should conduct thorough public consultations for at least 60 days on any proposed subsidiary legislation.
- Ensure continued flows of personal and other data. Given Taiwan’s export-oriented economy, it is vital to maintain an open legal framework that permits cross-border data transfers to facilitate international trade and investment.
- Promote the use of anonymized and de-identified data. In line with the PDPA’s definition of personal data as “information that may be used to directly or indirectly identify a natural person,” we recommend that data rendered truly anonymous, with no possibility of re-identification, not be classified as personal data under the PDPA. Such clarity would reduce compliance burdens and support innovation across sectors. We also recommend that the government adopt global anonymization and de-identification standards in future PDPA amendments, thereby encouraging innovation in privacy-enhancing technologies.
Suggestion 2: Strengthen government cybersecurity and procurement policy through sustainable investment and alignment with international standards.
- Ensure adequate and sustained cybersecurity funding to support full implementation of ZTA. According to the National Development Council’s initiative to promote the information security industry, cybersecurity funding should account for 5% to 7% of total IT expenditures, with this benchmark maintained or increased beyond 2025. We recommend that the government meet this target by increasing or reallocating funds to ensure full ZTA deployment.
- Align government cloud service procurement with international security and data protection standards. Under the joint government procurement contract for cloud services, the “Reference List of Basic Requirements for Common Information and Communication Security” is applied to cloud services before listing. However, the List’s “Application Software or System Development Services” category requires suppliers to submit security documentation and testing for inspection. This requirement is overly broad and should apply only to custom-developed solutions, not to public cloud services with standardized features used across clients.
Requiring public cloud providers to disclose proprietary source codes raises serious concerns regarding trade secrets and fails to meet principles of necessity and proportionality. It also creates uncertainty, as the List does not clearly define the scope of application software or system development.
We recommend discontinuing the current pre- inspection requirement. Instead, the government should assess whether cloud services meet widely recognized international or industry standards (including ISO/IEC 27001, ISO/IEC 27018, SOC 2). This approach would reflect best practices in global cloud procurement and enable agencies to adopt secure, compliant cloud services more efficiently. - Base ICT procurement decisions on international cybersecurity standards rather than country of origin. Under Executive Yuan Document No. 1090201804A, ICT products used by government agencies, including software, hardware, and services, must not originate from mainland Chinese brands. In 2024, the Executive Yuan approved a draft amendment to the Cyber Security Management Act, elevating these restrictions into law under Article 11, which prohibits public agencies from procuring or using products that pose national ICT security risks.
Focusing on country-of-origin, without clear technical criteria, may discourage international investment and complicate vendor evaluation. A standards-based approach ensures both security and openness. The Committee recommends shifting the procurement focus from country-of-origin to compliance with international standards for cybersecurity, personal data protection, and supply chain integrity (including ISO/IEC 27001, ISO/IEC 27701, ISO/IEC 20243, NIST SP 800-193, NIST SP 800-171). This approach still allows the government to exercise discretion in excluding non-compliant products, including those from mainland China, while offering a more transparent and consistent basis for evaluation. We also urge the government to identify a competent authority, such as the Ministry of Digital Affairs (MODA) or the Ministry of Economic Affairs, to address this critical issue.
To minimize procurement uncertainty and protect the rights of both vendors and agencies, we further recommend publishing the list of prohibited products referenced in Article 11-1 or creating a secure inquiry mechanism. In addition, the views of ICT manufacturers should be taken into consideration during the drafting of sub-laws and implementation measures to ensure practical enforcement.
Overreliance on country-of-origin restrictions, without clear technical guidelines, risks confusing procurement staff, deterring global companies, and weakening Taiwan’s ICT industry’s competitiveness in international supply chains. We urge the government to anchor future procurement policy in internationally accepted cybersecurity standards.
Suggestion 3: Ensure that AI regulation supports innovation through a principles-based, risk-proportionate framework.
To support Taiwan’s ambition of becoming an AI island under the Taiwan AI Action Plan 2.0, the Committee recommends that the AI Basic Act and related guidelines adopt a regulatory approach that promotes innovation while managing real risks. We offer the following recommendations:
- Adopt a principles-based framework and implement regulations grounded in a risk-based approach. AI regulation should focus on high-risk scenarios and avoid prescriptive regulation to maintain flexibility amid rapid AI technological advancements. Adopting a risk-based, principles-driven AI regulatory framework will not only encourage innovation but also attract global AI developers seeking a stable and trusted policy environment.
- Align with international standards. Considering Taiwan’s export-oriented trade model and global AI development trends, it is crucial to align with international or industry standards and practices (such as the U.S. NIST AI Risk Management Framework and ISO/IEC 42001). In accordance with the AI Basic Act, international cooperation should be actively promoted, such as with Taiwan’s AI Safety Institute (AISI), which facilitates cross-border collaboration on AI safety testing, research, and policy coordination to improve the global governance of advanced AI systems.
- Prioritize the application of existing regulations and incorporate industry opinions through public-private collaboration. New regulations, including reference guidelines, should only be established when current laws are insufficient, and industry opinions should be collected through public-private collaboration at an early stage.
- Tailor data protection requirements to application-specific risk levels. Regulations should support the use of AI by encouraging context-specific data protection measures, rather than blanket prohibitions. For example, bans on sharing confidential or personal data with generative AI services should distinguish between open models and closed, on-premises deployments. This risk-based approach enables practical use of AI while safeguarding sensitive data.
- Revise public sector guidelines to align with AI Basic Act principles. As MODA promotes AI applications across government agencies, the National Science and Technology Council (NSTC) reference guidelines on using generative AI should be updated to reflect the principles and risk management framework of the AI Basic Act. This alignment will support innovation, enable responsible AI adoption in the public sector, and help position Taiwan as a leader in AI-driven digital governance.
Suggestion 4: Initiate cross-agency efforts to build Taiwan’s quantum leadership through a multi-pronged approach.
- Expand R&D to accelerate quantum integration. Quantum computing is a revolutionary technology capable of solving complex problems beyond the reach of classical computers. It holds transformative potential across leading sectors such as material science, machine learning, cybersecurity, fintech, and transportation, while fostering new economic opportunities and high-value employment. The NSTC has identified quantum technology as one of its eight major forward-looking research platforms and is investing NT$8 billion (US$288.2 million) from 2022 to 2026 to advance this field.
We recommend that the government prioritize R&D in its next five-year plan by expanding resources for deploying and utilizing integrated quantum high-performance computing systems. These systems will significantly accelerate computational tasks for both scientific and industrial applications. Additional support should go toward research in quantum software and practical applications, including the identification of government specific-use cases. - Cultivate a quantum-ready workforce. Talent development is critical to meeting the growing demand for quantum technologies. The Committee urges the government to expand quantum education and training programs to cultivate a skilled domestic workforce. This includes academic partnerships, curriculum development, and training opportunities across all levels of education and professional development.
- Incentivize private-sector engagement. Industry participation must be a core pillar of Taiwan’s national quantum strategy. The framework should encourage involvement from both large enterprises and emerging startups through financial incentives, innovation grants, public-private research initiatives, and infrastructure support. This approach will help foster a robust talent pipeline and accelerate commercialization of quantum technologies.
- Develop a national strategy for post-quantum cryptography. One of the future risks of quantum computing is its potential to break current encryption standards. Taiwan should act now to integrate post-quantum cryptography into national security and industry protection plans. To prepare Taiwan’s digital infrastructure for the coming quantum era, we recommend the development of a national quantum readiness plan focused on post-quantum cryptography (PQC). This plan should incorporate PQC into existing cybersecurity strategies and encourage early migration to globally recognized post-quantum algorithms. Doing so will ensure Taiwan’s long-term data security and international competitiveness in cybersecurity resilience.
Suggestion 5: Enable Taiwan’s digital future by ensuring fair treatment for internet data centers and cloud service providers.
Internet data centers (IDCs) and cloud service providers (CSPs) are critical to Taiwan’s digital economy, supporting a significant share of its economic activity and enabling digital transformation, technological leadership, and national resilience. As major customers of Taiwan’s semiconductor industry and essential enablers of AI development, IDCs and CSPs reinforce Taiwan’s strategic position in the global technology supply chain and as a regional AI hub. IDC establishment also drives large-scale infrastructure investment, including in fiber optic networks and transport systems, leading to direct and indirect job creation across construction, engineering, and services, with significant contributions to Taiwan’s GDP.
- Ensure fairness in electricity pricing for IDCs. Electricity is the largest operating expense for IDCs and CSPs, which run high-availability digital infrastructure 24/7. However, recent electricity price hikes have created a growing cost burden, threatening the competitiveness of these sectors. In April 2024, Taipower introduced Schedule 6, imposing an additional price adjustment on IDCs whose annual electricity consumption exceeds 50 million kWh. This measure, which singles out IDCs from other growing industries classified under Schedule 5, is unprecedented in the Asia-Pacific region and lacks a clear rationale.
The Committee urges the government to revise the electricity pricing structure to uphold principles of fairness and non-discrimination. In particular, IDCs should not be singled out as a distinct pricing category. Instead, pricing policies should apply consistently across all industries, with classifications based on objective criteria such as voltage levels, load profiles, and grid utilization. A holistic approach will help prevent the disproportionate allocation of Taipower’s cost recovery burden onto a single sector, and ensure a more equitable and sustainable electricity market. We recommend removing Schedule 6 and reclassifying IDCs under Schedule 5 to restore parity and promote balanced economic development. - Promote transparency and consultation in electricity pricing policy. The classification process and calculation of electricity price adjustments must be transparent and inclusive. Effective policy should involve early consultation with relevant industry stakeholders to prevent unintended economic consequences. A well-defined, predictable mechanism for electricity pricing will provide long-term visibility and stability, enabling IDCs and CSPs to make infrastructure investments with confidence. A clear, consultative pricing policy is essential for Taiwan to maintain its attractiveness as a regional digital infrastructure hub.
- Implement performance-based energy standards for CSPs. Ongoing discussions on potential new compliance requirements for CSPs, such as mandatory disclosures on technical and efficiency metrics, risk undermining Taiwan’s technological competitiveness and cloud infrastructure resilience. The Committee urges the government to recognize the inherent operational efficiencies of cloud services, which benefit from dynamic workload optimization, high utilization rates, and advanced cooling technologies.
We recommend adopting performance-based energy standards rather than prescriptive technical requirements. Metrics like Power Usage Effectiveness remain useful, but target values should reflect facility type, operating patterns, and local climate conditions. Regulatory frameworks should focus on environmental impact and overall energy efficiency, allowing CSPs the flexibility to pursue innovative solutions that may exceed traditional efficiency benchmarks.
本委員會高度讚賞並支持政府推動的「資安即國安2.0」策略,該策略強調導入零信任架構(ZTA)的重要性,以實現可信賴的驗證程序和更強化的數位保護。然而,政府2025年的資安預算遭削減,對全面實施零信任架構造成挑戰。預算充足對於建置有效的身份驗證機制、設備認證和信任推論至關重要。
我們建議政府適度額外編列資源或調整現有預算,以強化資訊安全保護措施。為配合「臺灣AI行動計劃2.0」將台灣轉型為「AI之島」的國家目標,我們建議採用基於原則性的框架,在促進AI創新與風險管理之間取得平衡,該框架應符合國際標準,並優先考慮公私部門的合作。
委員會亦建議啟動跨部會合作,以積極提升台灣在全球量子計算領域的競爭力。量子計算可帶動台灣的關鍵產業的革新,並大幅提升其競爭力。本委員會建議優先制定一項新的五年計劃,聚焦於增強研發投入、培養人才、激勵產業參與,以及制定後量子密碼學的策略。
本委員會鼓勵持續相關部會的對話和合作,共同加強台灣的科技韌性與資安上的整體能力。
建議一:現代化台灣個人資料保護法規,平衡經濟發展與個資維護
本委員會肯定個人資料保護委員會籌備處(下稱「個資籌備處」)為於今年八月前設立個資會所投注之努力。在此籌備過程中,個資籌備處並就《個人資料保護法》(下稱「個資法」)提出修正草案。本委員會建請善用此契機,建構符合時宜之個資保護法制,同時審慎衡量台灣經濟結構之轉型及對外貿易的高度依賴。具體建言如下:
- 釐清資料控管者與資料處理者之區別
現行《個資法》第四條並未明確區分資料控管者與資料處理者之角色及責任。為與國際標準接軌並釐清法律責任,本會建議明確將資料控管者定義為決定處理個資之目的及方法者,而資料處理者則定義為代表資料控管者處理個資者。
資料處理者之義務應限於執行適當之安全維護措施,並遵循其與資料控管者間契約所約定之事項。由於資料處理者通常缺乏與資料當事人有直接關係及對於資料脈絡之掌握,因此不宜賦予其與資料控制者相同的法律責任。
明確區分兩者角色將有助於提升問責性與法遵程度,並促使臺灣個資保護制度與國際規範接軌。儘管此議題已於2024年白皮書中提出,至今尚無實質進展。
因此,不宜賦予其與資料控制者相同的法律責任。明確區分兩者角色將有助於提升問責性與法遵程度,並促使臺灣個資保護制度與國際規範接軌。儘管此議題已於2024年白皮書中提出,至今尚無實質進展。 - 賦予資料控管者通報個資外洩通報義務
承接建議1.1所提之角色區隔,《個資法》應明確規定,資料控管者而非資料處理者,負有向主管機關通報個資外洩事件及通知受影響當事人之責任。資料處理者僅需將個資外洩事件通知給資料控管者。上述義務應俟相關子法訂定完成後始生效力,該子法應明確規範通知通報門檻、機制及時限等事項,以免衍生法規遵循之疑義。 - 以損害門檻作為當事人通知之啟動條件
一套有效之個資外洩通知機制,應以告知當事人實際風險,並提供降低潛在損害所需資訊為目的。參考歐盟《一般資料保護規則》(GDPR)及新加坡《個人資料保護法》之規範,當事人通知義務應設立損害門檻。此舉可避免過度通報之情形,因為頻繁通知輕微或無實質威脅之事件,恐致當事人對重要警示產生疲乏感,進而削弱重要警示之效益。 - 違反通報及紀錄保存規定者,應先給予改正期間再行處罰
最新之《個資法》修正草案對於違反通報及紀錄保存等行為,未給予改正期間即逕予處罰。然此類違規情形往往非屬故意,且組織機構通常需要時間評估外洩事件之性質及影響,以釐清其法遵義務。新興數位產業也可能因主管機關管轄權認定不明,而產生非故意之遲延。本會建議比照個資外洩事件通知當事人之作法,對於違反通報及紀錄保存規定者,亦應設置改正期間。 - 採行彈性化的安全維護機制
主管機關於訂定安全維護要求及管理機制時,應避免過度細瑣之規範,並允許非公務機關得依其產業特性、風險程度及所處理個資之類型,彈性採行安全維護措施。相關子法應支持組織機構得依其個別情境及營運需求,採行符合比例原則且風險導向之彈性措施。 - 設置至少兩年之過渡期,以利新版《個資法》之法遵作業
非公務機關,尤其是營運規模大、業務較為複雜之企業,須有充分時間建置相關流程及機制,以符合個資保護新制之要求,尤其是個資外洩相關規範。為確保有效法遵,本會建議於新制公布日後,設置至少兩年之過渡期。 - 針對個資法子法草案進行完整公眾意見諮詢
最新《個資法》修正草案賦予主管機關廣泛之法規命令訂定權限,得以研擬施行細則。為確保所有利害關係人均有充分且實質的機會參與政策制定,個人資料保護委員會對於任何子法草案應辦理至少六十日之完整公眾意見諮詢。 - 確保個人資料及其他資料得以持續跨境流通
鑒於台灣經濟係以出口導向為主,維持開放性之法制架構,允許跨境資料傳輸,對於促進國際貿易與投資至關重要。 - 推動匿名化與去識別化資料之運用
依《個資法》對個人資料之定義:可直接或間接方式識別該個人之資料,本會建議將已完全匿名化且無從重新識別之資料,排除於《個資法》規範之列。此一明確性可減輕法遵負擔,並支持各產業創新發展。本會亦建議政府未來修法時,參採國際通用之資料匿名化與去識別化標準,進而鼓勵隱私強化技術之創新應用與發展。
建議二: 透過永續投資與接軌國際標準,強化政府資安與採購政策
- 保障充分且持續的資安預算,以支持全面落實零信任架構
根據國家發展委員會所出版的《台灣經濟論衡2024年3月號》中指出因應產業轉型所需資安防護,推動資安卓越產業發展方案,各政府計畫資安經費編列占資訊總經費 5%至 7%以上,自2021年起配合政策逐年調整,至2025年後占計畫總經費 5%至 7%以上。
本委員會建議政府透過增加或重新分配資安預算,以達成此一目標,並確保零信任架構得以全面實施。 - 政府雲端服務採購應採用比照國際資安和資料保護的標準
依據《雲端服務共同供應契約》(以下簡稱「雲端共契」),要求雲端品項上架前須先適用《各類資訊 (服務) 採購之共通性資通安全基本要求參考一覽表》(以下簡稱「採購參考表」)。然而,《採購參考表》中的「應用軟體或系統開發服務」類型,要求廠商提供原始碼送測。此要求範圍過於廣泛,應僅適用於客製化開發之應用軟體或系統,而不應延伸至對所有客戶提供標準化功能的公共雲端服務。要求公共雲端廠商揭露其專有原始碼,恐引發營業秘密外洩的風險,也有違必要性及比例原則。此外,該表未界定「應用軟體或系統開發服務」之適用範圍,導致雲端服務上架時須面臨高度不確定性。
本委員會建議不再對雲端品項進行上架前檢測,而改檢閱雲端服務是否符合國際或產業認可標準(例如:ISO/IEC 27001、ISO/IEC 27018、SOC 2)。此舉反應全球雲端服務採購之最佳實作,使政府機關更有效率地採購更多符合資安與法規要求的雲端產品。 - 政府資通訊產品採購政策應以國際資安標準為依據,而非產品來源國
依據行政院2020年12月18日發布之院臺護長字第1090201804A號函,目前政府規定公務用之資通訊產品(含軟體、硬體及服務)不得使用中國大陸廠牌。為強化國家整體資通安全之法律規範,行政院亦於2024年通過《資安管理法修法草案》新增第11條規定,明文禁止公務機關採購及使用可能危害國家資通安全之產品,將既有限制使用原則提升至法律位階。
若僅專注於產品原產地,而未明確界定技術標準,恐會阻礙國際投資,並使供應商評估變得更複雜。以標準為基礎的作法,則能兼顧資安與開放性。本委員會建議政府應將資通訊產品採購重點,從產品原產地轉向基於國際資安、個資保護及供應鏈安全等相關標準(例如:ISO/IEC 27001、ISO/IEC 27701、ISO/IEC 20243、NIST SP 800-193、NIST SP 800-171)。此一做法仍賦予政府排除不合規產品(包括來自中國大陸的產品)之裁量空間,同時提供一個更透明且一致的評估基準。我們也敦促政府明確指定主管機關(如數位發展部或經濟部)統籌處理此關鍵議題。
為降低採購不確定性,並保障供應商與相關政府機關雙方權益,本委員會進一步建議,政府應公布目前院版草案第11條之1所提及的禁用產品清單,或建立一套安全的查詢機制。此外,在制定《資安法》相關子法與實施細則時,應納入資訊通信技術(ICT)製造商的意見,以確保實務上的可執行性。
過度依賴原產地限制,若缺乏明確的技術指引,將可能導致政府機關採購人員無所適從,並使國際企業卻步,進而削弱臺灣ICT產業在國際供應鏈中的競爭力。我們呼籲政府未來的資通訊產品採購政策,應以國際公認的資安標準為基礎。
建議三:人工智慧(AI)規範應透過原則導向、風險比例適洽之框架,支持創新發展
為協助實現「臺灣AI行動計畫2.0」打造「AI臺灣島」的國家目標,本委員會建議《人工智慧基本法》及相關規範,應採取兼顧創新發展與風險管理之監理架構,我們建議採下列建議為之:
- 採框架式及原則性導向之規範,且依風險程度調整監理強度
AI監管應著重於高風險應用情境,並避免細瑣性規範(Prescriptive-based Approach),以保有因應技術快速演進之彈性。採行以風險為基礎、原則導向之AI監理架構,除能鼓勵創新外,亦有助於吸引尋求穩定與可信政策環境的全球AI開發者。 - 與國際標準接軌
考量臺灣輸出導向之外貿模式及全球AI發展趨勢,與國際或產業標準與實作接軌 (例如:美國NIST 《AI風險管理框架》及ISO/IEC 42001等規範)至關重要。依據《人工智慧基本法草案》,應積極推展人工智慧國際合作,例如:透過臺灣人工智慧安全研究院(AISI)參與跨國協作,針對AI安全測試、研發與政策協調進行合作,以強化全球先進AI系統之治理能量。 - 優先適用現有規範,並透過公私協作納入產業意見
新規範(包括參考指引)應僅於現行法令不足以因應時始予制定,且應於初期透過公私協作蒐集產業意見 。 - 根據特定使用情境的風險程度制定資料保護規定
法規應採用依特定使用情境的資料保護措施來支持人工智慧運用,而非概括禁止特定技術或系統。採用依特定情境的風險基礎方式,可使人工智慧得以被運用,同時保護敏感性資料。 - 修訂公部門指引,使其與《人工智慧基本法》的原則一致
隨著數位發展部在各政府機關推動AI應用,國家科學及技術委員會針對生成式AI使用所訂定的參考指引,應進行更新,以反映《人工智慧基本法》所揭示的原則與風險管理框架。此對照將有助於推動創新、促進公部門負責任地採用AI,並協助臺灣在AI驅動的數位治理方面成為領導者。
建議四:啟動跨部門合作,透過多元策略推動臺灣量子科技的領導地位
- 擴大研發投入,加速量子技術整合應用
量子計算是一項革命性技術,能解決傳統電腦難以或無法處理的複雜問題。量子計算在材料科學、機器學習、資訊安全、金融科技與交通運輸等關鍵領域具有顛覆性潛力,並有助於創造新興經濟產業與高附加價值的就業機會。國家科學及技術委員會已將量子科技列為八大前瞻科研平台之一,並於2022年至2026年間投入新臺幣80億元(約2.9億美元)發展量子相關研究。
我們對國科會的前瞻性規劃表示肯定。隨著量子優勢時代預期即將在未來幾年內來臨,我們建議政府在下一期五年計畫中,將研發列為優先項目,並擴大資源投入於整合型量子電腦與高效能計算系統的部署與應用。這些系統能大幅加速科學與產業應用中的運算任務。此外,政府亦應支持量子軟體與應用的研究,包括針對政府部門的應用場景進行探索。 - 培育具備量子才能之本地人才
面對量子科技日益增長的需求,人才發展至關重要。本會建議政府擴展量子教育與訓練計畫,以培養具備實務技能的國內量子人才,涵蓋產學合作、課程設計與各級教育及專業培訓機會。 - 鼓勵民間產業參與
民間產業參與應成為臺灣量子國家戰略的核心支柱之一。此一政策架構應透過財政獎勵、創新補助、公私合研計畫與基礎設施支持,鼓勵大型企業與新創公司共同投入。此舉有助於建構健全的人才培育管道,加速量子技術之商業化進程。 - 制定全國性後量子密碼學策略
量子運算的潛在風險之一,是其可能破解現有加密標準。臺灣應立即行動,將後量子密碼學(PQC)整合進國家安全與產業資安防護計畫中。為迎接即將到來的量子時代,我們建議政府擬定聚焦於 PQC 的全國性量子準備計畫,將PQC納入現有資安策略,並鼓勵及早採用全球公認的演算法進行轉移。此舉將有助於確保臺灣長期的資料安全與國際資安韌性競爭力。透過建立協作生態系統,這些努力將推動創新、強化國家安全,並釋放量子運算對社會的轉型潛力。
建議五:確保提供網路資料中心與雲端服務供應商公平的營運環境,以賦能台灣的數位未來
網路資料中心 (IDCs) 與雲端服務供應商 (CSPs) 對於台灣數位經濟體系至關重要,不僅支撐著相當比例的經濟活動,更在推動數位轉型、確立技術領導地位及增強國家韌性方面扮演關鍵角色。憑藉其作為台灣半導體產業主要客戶以及人工智慧 (AI) 發展核心賦能者的雙重角色,資料中心與雲端服務供應商強化了臺灣在全球科技供應鏈中的戰略地位,以及其作為亞太地區人工智慧樞紐的地位。資料中心的建置亦驅動大規模基礎建設投資,範疇涵蓋光纖網路與運輸系統等,進而在營造、工程及服務等產業創造直接與間接的就業機會,對國內生產總值 (GDP) 貢獻卓著。
- 確保網路資料中心電費計算方式之公平性
資料中心與雲端服務供應商需維持全年無休運作高可用性的數位基礎設施,因此電力成本通常為其最主要的營運支出。然而,近期的電價上漲已對此產業構成日益沉重的成本負擔,並威脅其整體競爭力。
台灣電力公司於2024年4月新增第六類電價計費方式。針對當年度用電量在 0.5 億度(kWh)的資料中心加收額外電價。此舉將網路資料中心產業單獨排除自第五類成長型產業的電費計算方式(Schedule 5)之外,在亞太地區前所未見,且缺乏明確的理據。
本委員會強烈建議政府重新審視現行電價結構,以落實公平與不歧視原則。資料中心不應被單獨列為獨立的電價分類,電價政策應一體適用於各產業,其分類應基於電壓等級、用電負載特性與電網利用率等客觀標準,整體性規劃有助於避免台灣電力公司將成本回收負擔不成比例地轉嫁至單一產業,並促進電力市場之公平與永續發展。為恢復產業間的平等並促進整體經濟的均衡發展,本委員會建議廢止第六類電力計費方式 (Schedule 6),並將資料中心產業重新歸類至第五類電費計算方式 (Schedule 5)。 - 強化電價政策制定的透明度與產業諮詢機制
電價調整的分類過程與計算方式,務必公開透明且具廣泛參與性。有效的電價政策制定,應在早期納入與相關產業利害關係人的諮詢,以避免未預期的負面經濟衝擊。一套完善且具備高度可預測性的電價機制,將為資料中心及雲端服務供應商提供長期穩定性與可預見性。有助於其安心推動基礎設施投資。明確且具產業參與機制的電價政策,亦是鞏固臺灣在亞太區域數位經濟領先地位的關鍵。 - 實施以績效為導向的能源標準
目前針對雲端服務供應商潛在新增的合規要求,例如強制揭露技術與效率指標,恐將有損台灣的技術競爭力與雲端基礎設施韌性。委員會敦促政府認可雲端服務內在的營運效率,其得益於動態工作負載最佳化、高利用率及先進冷卻技術。
我們建議對雲端服務供應商採納以績效為本的能源標準,而非規範式的技術要求。能源使用效率 (Power Usage Effectiveness, PUE) 等指標固然仍具參考價值,然其目標值應反映設施類型、營運模式及當地氣候條件。監管框架應著重於環境影響與整體能源效率,賦予雲端服務供應商彈性,以追求可能超越傳統效率基準的創新解決方案。
