Chunghwa Telecom recognizes its responsibilities with respect to the development of artificial intelligence (AI) and has established a responsible AI management system in accordance with ISO/IEC 42001.

Chunghwa Telecom incorporates the principles of security, trustworthiness, transparency, fairness/equity, accountability, and sustainability into the development, deployment, operation and maintenance, and continuous improvement processes of AI applications. This ensures that AI technologies are applied on a secure, compliant, and responsible basis, while continuously creating positive value for customers, the Company, and society.

Labeling Principles for AI-Generated Content and AI-Assisted Decision-Making

To enhance the transparency and trustworthiness of AI applications, Chunghwa Telecom has established a labeling mechanism for AI-generated content and AI-assisted decision-making, ensuring that users and stakeholders can clearly identify the extent of AI involvement and mitigating the risks of information misunderstanding and decision-making. For AI-generated content, or in scenarios involving analysis, recommendations, or decision support in external communications, documents, and customer service interactions, Chunghwa Telecom provides appropriate disclosure in line with the use context and explicitly labels such content or outcomes as “AI-assisted Decision-making,” while emphasizing that human personnel remain in the loop and responsible for final decisions.

Chunghwa Telecom integrates the labeling principles into its AI governance framework and application lifecycle management and adopts differentiated management measures in line with risk levels. For high-risk applications, the Company gives priority to strengthening disclosure and explanation mechanisms. Chunghwa Telecom also continuously reviews the transparency and explainability of AI applications through impact analyses and risk assessments, with the Artificial Intelligence Working Group responsible for supervision and review.

Taking the online customer service system of Chunghwa Telecom as an example, when customers access the service via the corporate website or an app, the system displays a conspicuous text label upon initiating AI interaction. The system proactively informs users that they are interacting with AI and explains the scope of real-time responses and auxiliary services provided by AI, ensuring that users can understand the role of AI involvement and the service limitations before the interaction begins.

In addition, during the development of AI applications, Chunghwa Telecom prepares and reviews DataCard and application inventory forms in accordance with its Software Development Procedures, confirming data sources, legality of use, and model-related information to enhance the transparency and explainability of AI applications. For AI-generated outputs, the Company also explicitly labels on the interaction page that such outputs are generated by AI.

AI Model Performance and Data Drift Monitoring and Correction Mechanism

To ensure that AI models maintain stable performance and security after deployment, Chunghwa Telecom promptly detects and corrects models when data distributions change or business rules are adjusted, preventing the escalation of erroneous decisions. The Company has integrated AI governance deep into its software development lifecycle, with enterprise-wide applicability, and implements the “AI Model Performance and Data Drift Monitoring and Correction Mechanism” to ensure service robustness and reliability.

Chunghwa Telecom implements enhanced controls over models used in internal and external systems, tracks model inputs and outputs, accuracy, and resource utilization in real time, and deploys automated guardrails to implement continuous monitoring and proactively detect data drift and model performance degradation. Upon detecting deviations in data distribution from the training baseline (i.e., data drift), accuracy falling below the predefined thresholds, or resource utilization spiking abnormally, the Artificial Intelligence Working Group assesses the risk and initiates model retraining and bias correction plans to ensure fair and accurate decision-making.

In addition, Chunghwa Telecom conducts a comprehensive audit at least once each year, coupled with an incident reporting mechanism to respond promptly to risks of operation and maintenance failure. With a “human-centric” responsible approach, the Company builds a secure and transparent digital trust environment and realizes its sustainable vision of Tech for Good.

The relevant mechanisms have been prioritized for introduction into the internal AI applications at Chunghwa Telecom and further extended to multiple digital transformation projects of government agencies. Meanwhile, the Company implements MLOps governance processes through Invent AI's AI PaaS (Platform as a Service) platform, optimizing management efficiency across the full model lifecycle.

AI Model Fairness and Bias Assessment Management Mechanism

To prevent systemic discrimination and safeguard digital human rights, Chunghwa Telecom has established a rigorous and regularly reviewed “AI Model Fairness and Bias Assessment Management Mechanism,” committed to eliminating discrimination and bias in technology development. The Company integrates fairness requirements deep into the software development lifecycle and regularly conducts AI risk assessments and impact analyses each year to identify potential bias at the source.

During the development phase, Chunghwa Telecom implements data diversity audits to ensure the creation of diverse and representative datasets and avoid demographic bias. The Company also regularly introduces quantitative monitoring metrics (e.g., demographic parity and equal opportunity) to detect and identify potential biases and combines these measures with multi-stage debiasing techniques to ensure that its decision-making adheres to the principles of fairness and non-discrimination.

Through a comprehensive management system, Chunghwa Telecom proactively prevents decision-making bias and the generation of inappropriate content, ensuring that AI services comply with social ethics and uphold human rights values, realizing its sustainable vision of “Tech for Good” and “Digital Equity.”

Chunghwa Telecom has independently developed a proprietary model evaluation tool, which provides comprehensive performance and compliance assessments for Large Language Models (LLMs). This mechanism has been widely applied in various smart service projects for government agencies. Through the tool, the Company regularly screens and assesses models to not only confirm that model outputs consistently align with expectations, but also strengthen risk controls and fairness evaluations, establishing a practical benchmark for responsible AI.

Initiatives to Lower the Ecological Footprint of AI Data Centers and Models

Chunghwa Telecom has systematically lowered the ecological footprint of AI data centers through measures such as high-efficiency data center design, renewable energy adoption, a carbon inventory system, and AI computing optimization. These measures have also been incorporated into the ISO and ESG disclosure systems at the Company, supported by auditable evidence.

1. Policy and Governance: Chunghwa Telecom has incorporated “lowering the environmental impact of data centers and AI computing” into its ESG and operational governance framework and established a comprehensive management system：
   • Introduction of ISO 50001 (Energy Management)
   • Introduction of ISO 14001 (Environmental Management)
   • Establishment of carbon reduction and energy efficiency KPIs, which are incorporated into the disclosure in the annual sustainability report
2. Technical Measures (Data Center and AI Efficiency): Chunghwa Telecom has adopted multiple energy-saving measures at the data center and AI computing levels：
   (1). Data Center Level
   • Avoiding overdesign, optimizing the number of servers, establishing hot and cold aisles, and implementing effective airflow management;
   • Increasing the operating temperature of server rooms to reduce air-conditioning power consumption and installing blanking panels in empty rack spaces to prevent short-circuiting of cold air;
   • Decommissioning or consolidating energy-intensive ICT equipment rooms;
   • Consolidating or shutting down legacy equipment, and adopting high-efficiency, low-energy-consumption, and high-temperature-tolerant ICT equipment;
   • Adopting high-efficiency electrical equipment;
   • Replacing legacy air-conditioning units with high sensible heat air conditioners, and replacing water chillers with high-efficiency magnetic-bearing variable-frequency chillers;
   • Replacing legacy water pumps with ultra-high-efficiency water pumps rated IE3 or above;
   • Adding automatic water quality maintenance systems to improve the efficiency of cooling water systems;
   • Introducing AI-driven smart energy-saving technologies (intelligent optimal operating mode);
   • Prioritizing liquid cooling solutions for future AI data center (AI DC) developments, while continuously introducing innovative approaches (such as immersion cooling).
   (2). AI Computing Level：
   • Virtualization and resource scheduling to enhance equipment utilization;
   • Avoid the inefficiencies of fragmented or under-utilized distributed computing.
   (3). Model and Algorithm Level：
   • Monitoring model performance to reduce additional computing caused by performance degradation;
   • Tracking resource utilization to improve the efficiency of computing resources used by models;
   • Performing model tuning and retraining to minimize redundant or inefficient computing;
   • Improving the computational efficiency of AI models to reduce energy consumption and the carbon footprint of models
3. Energy and Carbon: Chunghwa Telecom actively promotes renewable energy and carbon management：
   • Procuring renewable energy (including green electricity and RECs) – The proportion of renewable energy has increased year by year, with the target for the IDC green power wheeling rate raised to 42% in 2026, accelerating progress toward the target of 100% renewable energy for IDCs by 2030 (RE100);
   • Establishing a carbon inventory system (compliant with ISO 14064) – 100% inventory coverage for Scope 1 and Scope 2 emissions;
   • Regular external disclosure (ESG reports and CDP).
4. Operation Management: Chunghwa Telecom ensures continuous optimization through systematic management：
   • Deploying the DCIM/iEN systems to monitor power and cooling systems in real time;
   • Conducting regular energy audits and improvement plans;
   • Adopting the PDCA cycle for energy efficiency management.
5. Sustainability Impact: Through the aforementioned measures, Chunghwa Telecom has achieved results as follows：
   • Reducing the operational carbon footprint of data centers.
   • Improving energy use efficiency and reducing operating costs.
   • Complying with the ESG requirements of international cloud and AI customers.
   • Elevating corporate sustainability ratings and international competitiveness.

Appeal and Remediation Mechanisms for AI Decisions

To ensure the fairness and accountability of artificial intelligence applications, Chunghwa Telecom leverages its current customer complaint reporting and handling system. Through existing reporting and customer service mechanisms, the Company processes AI-related appeals and has instituted a standardized handling procedure, including case acceptance, classification and analysis, human review, and result notification, thereby establishing the appeal and remediation mechanism for AI decisions.

In accordance with its existing customer complaint handling process, Chunghwa Telecom clearly outlines the handling mechanism and prescribed processing timeline to complainants upon case intake. In principle, an initial response for general cases must be completed within 3 working days. For cross-unit or complex cases, the Company proactively informs the complainant of the extended timeline and estimated completion date, assigns a case number, and continuously reports progress via phone, SMS, apps, or online customer service across stages of “Acceptance,” “Processing,” and “Response/Closure.”

Upon case closure, Chunghwa Telecom provides the handling results and explanations. Where necessary, dedicated personnel will be arranged to provide further communication and clarification or to take remedial actions. Through explicit timeline management, progress disclosure, and response mechanisms, the Company ensures that customers can promptly capture the handling status and that their concerns are properly addressed.

For cases involving personal rights and interests or high-stakes decisions, human intervention and decision reassessment mechanisms are prioritized to ensure the fairness and reasonableness of decisions.

When biases or adverse impacts in AI decisions are detected, the Company takes appropriate remedial measures, including outcome correction, process improvement, and model optimization. Relevant cases are also incorporated into the risk management and continuous improvement mechanisms.

Through the aforementioned mechanisms, Chunghwa Telecom strengthens the appealability and accountability governance of AI, ensures that the AI applications comply with human-centric principles, and establishes a trustworthy digital service environment.

In addition, to address the responsibilities, risks, and governance issues potentially arising from AI adoption, Chunghwa Telecom actively strengthens its professional capabilities in AI governance and supports employees in obtaining ISO/IEC 42001 Lead Auditor certification.

Through systematic training and certification, Chunghwa Telecom has cultivated core personnel with capabilities in the planning, implementation, auditing, and continuous improvement of AI management systems. This effectively underpins the Company's initiatives in advancing AI governance, risk control, and institutional development as well as helps establish management mechanisms compliant with international standards, further enhancing compliance management capabilities, governance maturity, and overall competitive advantage.

Chunghwa Telecom recognizes its responsibilities with respect to the development of artificial intelligence (AI) and has established a responsible AI management system in accordance with ISO/IEC 42001.

Chunghwa Telecom incorporates the principles of security, trustworthiness, transparency, fairness/equity, accountability, and sustainability into the development, deployment, operation and maintenance, and continuous improvement processes of AI applications. This ensures that AI technologies are applied on a secure, compliant, and responsible basis, while continuously creating positive value for customers, the Company, and society.

Labeling Principles for AI-Generated Content and AI-Assisted Decision-Making

To enhance the transparency and trustworthiness of AI applications, Chunghwa Telecom has established a labeling mechanism for AI-generated content and AI-assisted decision-making, ensuring that users and stakeholders can clearly identify the extent of AI involvement and mitigating the risks of information misunderstanding and decision-making. For AI-generated content, or in scenarios involving analysis, recommendations, or decision support in external communications, documents, and customer service interactions, Chunghwa Telecom provides appropriate disclosure in line with the use context and explicitly labels such content or outcomes as “AI-assisted Decision-making,” while emphasizing that human personnel remain in the loop and responsible for final decisions.

Chunghwa Telecom integrates the labeling principles into its AI governance framework and application lifecycle management and adopts differentiated management measures in line with risk levels. For high-risk applications, the Company gives priority to strengthening disclosure and explanation mechanisms. Chunghwa Telecom also continuously reviews the transparency and explainability of AI applications through impact analyses and risk assessments, with the Artificial Intelligence Working Group responsible for supervision and review.

Taking the online customer service system of Chunghwa Telecom as an example, when customers access the service via the corporate website or an app, the system displays a conspicuous text label upon initiating AI interaction. The system proactively informs users that they are interacting with AI and explains the scope of real-time responses and auxiliary services provided by AI, ensuring that users can understand the role of AI involvement and the service limitations before the interaction begins.

In addition, during the development of AI applications, Chunghwa Telecom prepares and reviews DataCard and application inventory forms in accordance with its Software Development Procedures, confirming data sources, legality of use, and model-related information to enhance the transparency and explainability of AI applications. For AI-generated outputs, the Company also explicitly labels on the interaction page that such outputs are generated by AI.

AI Model Performance and Data Drift Monitoring and Correction Mechanism

To ensure that AI models maintain stable performance and security after deployment, Chunghwa Telecom promptly detects and corrects models when data distributions change or business rules are adjusted, preventing the escalation of erroneous decisions. The Company has integrated AI governance deep into its software development lifecycle, with enterprise-wide applicability, and implements the “AI Model Performance and Data Drift Monitoring and Correction Mechanism” to ensure service robustness and reliability.

Chunghwa Telecom implements enhanced controls over models used in internal and external systems, tracks model inputs and outputs, accuracy, and resource utilization in real time, and deploys automated guardrails to implement continuous monitoring and proactively detect data drift and model performance degradation. Upon detecting deviations in data distribution from the training baseline (i.e., data drift), accuracy falling below the predefined thresholds, or resource utilization spiking abnormally, the Artificial Intelligence Working Group assesses the risk and initiates model retraining and bias correction plans to ensure fair and accurate decision-making.

In addition, Chunghwa Telecom conducts a comprehensive audit at least once each year, coupled with an incident reporting mechanism to respond promptly to risks of operation and maintenance failure. With a “human-centric” responsible approach, the Company builds a secure and transparent digital trust environment and realizes its sustainable vision of Tech for Good.

The relevant mechanisms have been prioritized for introduction into the internal AI applications at Chunghwa Telecom and further extended to multiple digital transformation projects of government agencies. Meanwhile, the Company implements MLOps governance processes through Invent AI's AI PaaS (Platform as a Service) platform, optimizing management efficiency across the full model lifecycle.

AI Model Fairness and Bias Assessment Management Mechanism

To prevent systemic discrimination and safeguard digital human rights, Chunghwa Telecom has established a rigorous and regularly reviewed “AI Model Fairness and Bias Assessment Management Mechanism,” committed to eliminating discrimination and bias in technology development. The Company integrates fairness requirements deep into the software development lifecycle and regularly conducts AI risk assessments and impact analyses each year to identify potential bias at the source.

During the development phase, Chunghwa Telecom implements data diversity audits to ensure the creation of diverse and representative datasets and avoid demographic bias. The Company also regularly introduces quantitative monitoring metrics (e.g., demographic parity and equal opportunity) to detect and identify potential biases and combines these measures with multi-stage debiasing techniques to ensure that its decision-making adheres to the principles of fairness and non-discrimination.

Through a comprehensive management system, Chunghwa Telecom proactively prevents decision-making bias and the generation of inappropriate content, ensuring that AI services comply with social ethics and uphold human rights values, realizing its sustainable vision of “Tech for Good” and “Digital Equity.”

Chunghwa Telecom has independently developed a proprietary model evaluation tool, which provides comprehensive performance and compliance assessments for Large Language Models (LLMs). This mechanism has been widely applied in various smart service projects for government agencies. Through the tool, the Company regularly screens and assesses models to not only confirm that model outputs consistently align with expectations, but also strengthen risk controls and fairness evaluations, establishing a practical benchmark for responsible AI.

Initiatives to Lower the Ecological Footprint of AI Data Centers and Models

Chunghwa Telecom has systematically lowered the ecological footprint of AI data centers through measures such as high-efficiency data center design, renewable energy adoption, a carbon inventory system, and AI computing optimization. These measures have also been incorporated into the ISO and ESG disclosure systems at the Company, supported by auditable evidence.

1. Policy and Governance: Chunghwa Telecom has incorporated “lowering the environmental impact of data centers and AI computing” into its ESG and operational governance framework and established a comprehensive management system：
   • Introduction of ISO 50001 (Energy Management)
   • Introduction of ISO 14001 (Environmental Management)
   • Establishment of carbon reduction and energy efficiency KPIs, which are incorporated into the disclosure in the annual sustainability report
2. Technical Measures (Data Center and AI Efficiency): Chunghwa Telecom has adopted multiple energy-saving measures at the data center and AI computing levels：
   (1). Data Center Level
   • Avoiding overdesign, optimizing the number of servers, establishing hot and cold aisles, and implementing effective airflow management;
   • Increasing the operating temperature of server rooms to reduce air-conditioning power consumption and installing blanking panels in empty rack spaces to prevent short-circuiting of cold air;
   • Decommissioning or consolidating energy-intensive ICT equipment rooms;
   • Consolidating or shutting down legacy equipment, and adopting high-efficiency, low-energy-consumption, and high-temperature-tolerant ICT equipment;
   • Adopting high-efficiency electrical equipment;
   • Replacing legacy air-conditioning units with high sensible heat air conditioners, and replacing water chillers with high-efficiency magnetic-bearing variable-frequency chillers;
   • Replacing legacy water pumps with ultra-high-efficiency water pumps rated IE3 or above;
   • Adding automatic water quality maintenance systems to improve the efficiency of cooling water systems;
   • Introducing AI-driven smart energy-saving technologies (intelligent optimal operating mode);
   • Prioritizing liquid cooling solutions for future AI data center (AI DC) developments, while continuously introducing innovative approaches (such as immersion cooling).
   (2). AI Computing Level：
   • Virtualization and resource scheduling to enhance equipment utilization;
   • Avoid the inefficiencies of fragmented or under-utilized distributed computing.
   (3). Model and Algorithm Level：
   • Monitoring model performance to reduce additional computing caused by performance degradation;
   • Tracking resource utilization to improve the efficiency of computing resources used by models;
   • Performing model tuning and retraining to minimize redundant or inefficient computing;
   • Improving the computational efficiency of AI models to reduce energy consumption and the carbon footprint of models
3. Energy and Carbon: Chunghwa Telecom actively promotes renewable energy and carbon management：
   • Procuring renewable energy (including green electricity and RECs) – The proportion of renewable energy has increased year by year, with the target for the IDC green power wheeling rate raised to 42% in 2026, accelerating progress toward the target of 100% renewable energy for IDCs by 2030 (RE100);
   • Establishing a carbon inventory system (compliant with ISO 14064) – 100% inventory coverage for Scope 1 and Scope 2 emissions;
   • Regular external disclosure (ESG reports and CDP).
4. Operation Management: Chunghwa Telecom ensures continuous optimization through systematic management：
   • Deploying the DCIM/iEN systems to monitor power and cooling systems in real time;
   • Conducting regular energy audits and improvement plans;
   • Adopting the PDCA cycle for energy efficiency management.
5. Sustainability Impact: Through the aforementioned measures, Chunghwa Telecom has achieved results as follows：
   • Reducing the operational carbon footprint of data centers.
   • Improving energy use efficiency and reducing operating costs.
   • Complying with the ESG requirements of international cloud and AI customers.
   • Elevating corporate sustainability ratings and international competitiveness.

Appeal and Remediation Mechanisms for AI Decisions

To ensure the fairness and accountability of artificial intelligence applications, Chunghwa Telecom leverages its current customer complaint reporting and handling system. Through existing reporting and customer service mechanisms, the Company processes AI-related appeals and has instituted a standardized handling procedure, including case acceptance, classification and analysis, human review, and result notification, thereby establishing the appeal and remediation mechanism for AI decisions.

In accordance with its existing customer complaint handling process, Chunghwa Telecom clearly outlines the handling mechanism and prescribed processing timeline to complainants upon case intake. In principle, an initial response for general cases must be completed within 3 working days. For cross-unit or complex cases, the Company proactively informs the complainant of the extended timeline and estimated completion date, assigns a case number, and continuously reports progress via phone, SMS, apps, or online customer service across stages of “Acceptance,” “Processing,” and “Response/Closure.”

Upon case closure, Chunghwa Telecom provides the handling results and explanations. Where necessary, dedicated personnel will be arranged to provide further communication and clarification or to take remedial actions. Through explicit timeline management, progress disclosure, and response mechanisms, the Company ensures that customers can promptly capture the handling status and that their concerns are properly addressed.

For cases involving personal rights and interests or high-stakes decisions, human intervention and decision reassessment mechanisms are prioritized to ensure the fairness and reasonableness of decisions.

When biases or adverse impacts in AI decisions are detected, the Company takes appropriate remedial measures, including outcome correction, process improvement, and model optimization. Relevant cases are also incorporated into the risk management and continuous improvement mechanisms.

Through the aforementioned mechanisms, Chunghwa Telecom strengthens the appealability and accountability governance of AI, ensures that the AI applications comply with human-centric principles, and establishes a trustworthy digital service environment.

In addition, to address the responsibilities, risks, and governance issues potentially arising from AI adoption, Chunghwa Telecom actively strengthens its professional capabilities in AI governance and supports employees in obtaining ISO/IEC 42001 Lead Auditor certification.

Through systematic training and certification, Chunghwa Telecom has cultivated core personnel with capabilities in the planning, implementation, auditing, and continuous improvement of AI management systems. This effectively underpins the Company's initiatives in advancing AI governance, risk control, and institutional development as well as helps establish management mechanisms compliant with international standards, further enhancing compliance management capabilities, governance maturity, and overall competitive advantage.