RTFP

The governance of content moderation at Temu is organised through a framework that integrates executive oversight with specialized operational functions. This framework is built upon six distinct yet deeply interconnected control domains and employs evidence-based and data-driven assessments to evaluate the design quality, operational and mitigation effectiveness of these controls. 1. Governance, Risk & Compliance Oversight At the highest level of this framework is Temu¡¯s Board of Directors who is responsible for approving and reviewing This ensures that platform safety and moderation issues are treated as core business priorities. In addition to the Board, Temu has established a Board-approved delegated decision-making structure for certain DSA matters. Central to this framework is Temu¡¯s DSA Compliance Function, led by the Head of the DSA Compliance Function who is directly appointed by the Board. This Function is independent of Temu¡¯s day-to-day operations and works closely with internal stakeholders, including the the Legal and Compliance Team (LCT) and the Trust and Safety Team (TST) to ensure that all standards and practices remain aligned with the DSA and reflect evolving regulatory expectations . 2. Policies & Standards Temu¡¯s governance framework is further underpinned by platform-wide policies and standards as well as the integration of policy development and operational enforcement on the platform. The TST and LCT collaboratively develop and improve these platform policies and standards, refining them based on feedback from legal counsel, industry experts, and regulatory updates. Key policies are hosted in a user-friendly Transparency Center for public access. 3. User Management & Onboarding Control This process serves as our first line of defence. Temu implements comprehensive onboarding and verification procedures to ensure accountability and prevent bad faith actors from accessing the platform. These measures are complemented by education programmes that guide users on their rights, responsibilities, and compliance obligations on the platform. 4. Detection & Enforcement At the heart of Temu¡¯s mitigation framework is a moderation system that combines automated detection with human review. This integrated approach proactively detects, intercepts and removes non-compliant product listings and content that violate Temu¡¯s rules and policies. The system also analyses user behaviour to flag potential bad-faith actors. To ensure accountability, Temu also employs a structured penalty scheme for traders. For repeated offenders and severe violations, we apply stringent measures such as service suspension and permanent blocklisting, effectively barring bad-faith actors from Temu. 5. User Rights and Redress Mechanisms Temu recognises that equal and transparent participation in the marketplace is a fundamental right. We ensure that every user affected by a platform decision¡ªsuch as a product removal or service suspension ¡ª has a direct avenue to be heard. Decisions are accompanied by a Statement of Reasons and a link to our internal appeals process. Furthermore, the framework provides clear guidance on seeking remedies through certified out-of-court dispute settlement bodies. 6. External Engagement Temu highly values feedback from external stakeholders and actively uses these insights to enhance its practices. By maintaining open and constructive dialogue, we continuously review and incorporate recommendations, as well as industry best practices, into our compliance framework to ensure ongoing improvement and responsiveness. During the reporting period, we actively participated in industry seminars and maintained active engagement with regulatory bodies and industry associations to ensure our awareness of, and responsiveness to, systemic risks on our platform.

Own-initiative content moderation on Temu is conducted through a combination of automated systems and human review processes designed to identify content that is potentially illegal or non-compliant with applicable laws or Temu¡¯s platform policies. Our automated systems are designed to proactively detect and moderate indicators of non-compliance. These systems utilise dozens of primary models¡ªincluding Optical Character Recognition (OCR) and semantic analysis¡ªalongside logic-based rules to analyse content such as product listings, advertisements and consumer reviews, across various risk sectors including Intellectual Property (IP), prohibited goods, and sensitive content. The actions that flow from automated detection depend on the system¡¯s confidence levels. For high-confidence violations, our systems trigger immediate enforcement actions, other potential violations are escalated for human review. For promoted listings, Temu applies additional ad-specific content moderation measures. Listings eligible for participation in promotions are subject to automated pre-promotion review to detect prohibited keywords, recalled items, and other indicators of illegal or incompatible content. For higher-risk categories or more complex cases, a secondary manual review is conducted based on continuously updated rules. Any subsequent modification to a promoted listing automatically triggers a renewed review prior to continued promotion. We also use automated systems to prevent previously banned traders from re-entering the platform under new identities. Enforcement actions follow a structured and graduated penalty scheme, which may include ad removal, product delisting, service suspension, permanent blocklisting, and other proportionate measures depending on the severity and recurrence of violations. Our human review processes complement the automated systems by assessing escalated edge cases. Supported by a dedicated and multilingual team of human moderators, we ensure that content is reviewed within the broader context of compliance with local laws, cultural sensitivities, and regulatory requirements. Measure of Exposure Exposure metrics are calculated based on logged-in Unique Visitors (UV). UV-based measurement is used as a reasonable proxy for distinct user exposure and helps mitigate against potential overcounting associated with repeated page refreshes, pagination, comparison behaviour, and search result reloads, which are actions inherent to the operation of an online marketplace like Temu. In the context of an online marketplace, where product listings are accessed primarily through user-initiated searches and comparisons rather than algorithmic content amplification, exposure metrics based on logged-in Unique Visitors provide an appropriate measure of how widely potentially violative content is made available across the user base. To assess product exposure, Temu differentiates between ¡°Informational non-compliance¡± and other product violations. The former category pertains to administrative or documentation deficiencies ¡ªsuch as missing manuals or labelling errors¡ªrather than inherent safety risks. While these listings remain subject to enforcement and visibility restrictions, they are excluded from core exposure metrics to provide a more accurate view of content with substantive safety implications. During this reporting period, the exposure rate for potentially infringing or non-compliant product listings (excluding ¡°Informational non-compliance¡± listings) was 5.08% of total product views in the EU, with an average of 444 UV per infringing/non-compliant listing. Regarding consumer feedback, the exposure rate for potentially non-compliant reviews was 0.04% of total review views in the EU, averaging 344 UV per non-compliant review.

Temu uses a comprehensive methodology to determine the total number of Full-Time Equivalents (FTE) human resources dedicated to content moderation, thereby ensuring the accuracy of reported data. The total figure reported represents a snapshot of the workforce as of 31 December 2025, and includes both internal employees and outsourced personnel who are actively engaged in content moderation. FTE Calculation Methodology To ensure accuracy, the human resources count is calculated on an FTE basis, with one FTE corresponding to one person working full time on content moderation activities. Internal and Outsourced Personnel For internal employees, the count includes teams whose primary daily operations consist of content moderation activities. Regarding outsourced staff, Temu maintains a comprehensive and centralised registry that tracks the number of outsourced moderators across different workstreams. These figures are collected through direct consultation with moderation team supervisors and receive final validation from the Head of the TST to ensure accuracy.

Human reviewers play a pivotal role in Temu¡¯s moderation ecosystem, providing the nuanced contextual judgement required for complex or ambiguous cases escalated by automated systems. To ensure high-quality decision making, we maintain high standards for all personnel involved in content moderation. Educational Qualifications and EQF Alignment All content moderators have relevant experience. For example, all full-time employees at Temu are required to hold a minimum of a bachelor's degree which corresponds to a level 6 under the European Qualifications Framework (EQF). This baseline ensures that our staff possesses advanced knowledge necessary for complex problem-solving in moderation. Specialized Expertise and Professional Recruitment Beyond general degree requirements, we actively recruit specialists for moderation workflows requiring high expertise. Our intellectual property (IP) team, in particular, is composed of senior experts and professionals with advanced backgrounds in IP law, data science, and information technology. Linguistic Proficiency and Regional Expertise We employ content moderators with proficient language skills across all of our content moderation teams, strategically deploying multilingual professionals to support our corresponding country and regional sites. These professionals are fluent in local languages such as English, French, German, Italian, and Spanish, ensuring wide coverage of the EU linguistic landscape. This linguistic depth allows our teams to identify and interpret cultural nuances and local slangs that are essential for accurate moderation. Furthermore, to address the linguistic diversity of the EU, moderators within these workstreams are required to hold recognized language certifications, such as a B2 level under the Common European Framework of Reference for Languages (CEFR). This requirement ensures they possess the necessary proficiency to accurately process and review content across various local languages. For further detail on the linguistic expertise and distribution of our human moderators, please see Tab 9 of our Transparency Report.

The accuracy of automated moderation measures on Temu is assessed through a methodological framework using three primary quantitative indicators: accuracy, precision, and recall. These metrics provide a comprehensive view of the reliability of our automated systems in identifying and acting upon infringing, potentially illegal, or non-compliant content. Accuracy measures the overall proportion of correct automated decisions, including both correctly identified violative content (listings and reviews) and correctly retained compliant content. Precision measures the proportion of automated removals that are confirmed to be correct, indicating the system¡¯s ability to avoid ¡®false positives¡¯. Recall measures the proportion of all violative content that are successfully detected by automated means. Methodology and sampling approach Across product listings and reviews, accuracy, precision, and recall indicators are derived from statistically representative samples covering all applicable official EU languages. In the case of reviews, the EU language attributions are derived from the automated detection of language attributes within the review texts. Sampling is conducted using statistically robust methods and reported with a 95% confidence interval to ensure the resulting indicators are representative of automated system performance across the service. Application of methodology to product listings For product listings, sampling is designed to reflect two major categories of violation types: IP-infringing listings and non-compliant listings. To ensure a comprehensive assessment of the performance of automated systems, we use a stratified sampling methodology. Input data is drawn from multiple operational content pools representing different stages of the moderation lifecycle: listings subject to direct automated enforcement; listings not captured by direct automated enforcement that are subsequently enforced through moderator intervention loops; and listings that remain unenforced. Drawing from these diverse pools enables the platform to establish a robust baseline for evaluating both the accuracy of automated enforcement actions and the extent to which automated systems identify relevant violations. Ground truth is established through the manual review of sampled listings across these pools. The indicators are then calculated using a weighted extrapolation methodology that scales sample-level outcomes to platform-level populations. Application of methodology to consumer reviews While the core methodology remains consistent, we have adapted the parameters to reflect the distinct linguistic and behavioural characteristics of review content. To ensure a comprehensive view, we employ a random sampling of reviews across the EU platform to provide a representative overview of all marketplace activity, including reviews without text. As with product listings, input data for reviews is then drawn from multiple operational pools representing different stages of the moderation lifecycle, with ground truth established through manual review of moderation outcomes across these pools. The indicators are then calculated based on the result of manual review.

Temu utilises automated moderation technology to detect and moderate suspected illegal or violative content quickly and at scale. Our approach is governed by a strict and evolving set of parameters ensuring proportionality and accuracy. General Operational Parameters for Automated Moderation Our automated systems are governed by parameters, including the following: 1. Confidence Thresholds: Our systems assign a confidence level to every detection. High confidence matches trigger immediate enforcement, while lower-confidence cases are escalated for human review to ensure contextual accuracy. 2. Heuristic Feedback Loops: Automated systems are continuously refined using insights from manual review to reduce both false positives and false negatives. Area Specific Parameters for Automated Moderation 1. Trustworthy Reviews We know how important reviews can be when shopping online, so we are committed to ensuring reviews on Temu are authentic, reliable, and free from inappropriate or illegal content. Our tools scan and analyse reviews before and after publication to identify violative content, such as inappropriate terms and phrases, ensuring that such reviews are either removed or prevented from being published. Clearly violating detected reviews are automatically removed, and ambiguous cases are escalated for human review. For complex content, such as multilingual text or content embedded in images and videos, we employ machine translation and algorithms to help identify violations across different formats and languages. 2. Trader Onboarding and Verification We also use algorithm-based screening tools to scrutinise information provided by traders as part of their onboarding process, including business registrations and identification documents. In conjunction with human review, these tools help us to identify applicants who provide unreliable, incomplete or inaccurate information. 3. Product Safety and Compliance Prior to and after being posted on Temu, all product listings are screened by algorithm-based tools utilising advanced text, image, and video recognition algorithms. These systems analyse product data, including descriptions, images, labelling information, and qualification documents, against pre-set requirements and standards for each product category. This process identifies potential risks associated with prohibited, hazardous, or non-compliant products. If the system is not able to precisely identify whether a product is violative, the case will be referred for human review. Complementing this, we also deploy an automated cross-check mechanism to identify and take action against products similar to those previously removed for being violative. 4. Brand Protection To mitigate the risk of IP infringement, we use advanced technologies to conduct regular, comprehensive scans of listings on Temu. The system parameters are set to identify products by analysing text, logos and images to determine if they are identical or significantly similar to IP rights in our database, thereby indicating a high likelihood of IP infringements. This includes matching technology that cross-references listing titles and descriptions against protected trademarks to facilitate proactive identification and moderation by our IP content moderators. 5. Ad-specific Content Moderation Promoted listings are subjected to moderation procedures precisely tailored for ad-related compliance. The first layer is automated and utilises text and image recognition to detect sensitive content, prohibited keywords or recalled items within product titles and descriptions. Items that cannot be conclusively assessed by the automated system are reviewed manually.

Temu is committed to ensuring the reliability and effectiveness of its automated moderation systems. Our safeguard architecture is fundamentally grounded in the ¡°human in the loop¡± principle, ensuring continuous human oversight throughout the process. Pre-implementation of Automated Tools and Continuous Validation Prior to deployment of any algorithm or model, we conduct rigorous testing to ensure the tool¡¯s capability to identify violative content with a stabilized, high level of accuracy. Automated tools are only integrated into our content moderation system once they meet these strict performance benchmarks. Once the automated measure is live, we monitor daily operational data to assess and optimise performance and identify abnormalities. Detected irregularities are analysed to identify root causes and implement improvements. We also carry out regular checks on automated measures, including evaluations by specialized staff, to assess the accuracy of automated decisions. Safeguards accompanying our automated tools Where the automated systems flag content as potentially problematic, such content then undergoes a secondary review by our human moderators. 1. Trustworthy reviews All our reviews are scanned by automated tools to detect content that violates our rules. To ensure no violative content is missed, we supplement this with a manual reporting function for both users and traders. Every report is reviewed by our content moderation team, and appropriate action is taken. At the scale and complexity we operate, we ensure that any consumer can appeal our decision, where they disagree with our decision to restrict content. 2. Trader onboarding and verification We use automated means to screen and validate traders prior to listing. Any inconclusive screening results are escalated for manual review, with human reviewers cross-checking trader information and supporting documents against official online databases and third-party platforms to verify reliability and completeness. Where verification fails following human review, relevant traders may appeal the decision, or update information and attempt the verification process again. 3. Product safety and compliance Product listings that our automated technology cannot immediately categorize as violative or compliant are escalated for manual review. Furthermore, all user reports submitted via the ¡°Report this item¡± button are routed directly to human moderators, bypassing automated enforcement to ensure nuanced assessments. We also apply a combination of automated and manual measures to detect and sanction repeated violations by traders. Where violations are confirmed, a range of sanctions are imposed, such as warnings, removal of non-compliant products, or removal of all products from the store. However, traders on Temu always have the option to contest our moderation decisions by appeals. 4. Brand protection Following automated detection of ambiguous IP infringements, specialized multilingual moderators with extensive IP experience review the listing¡¯s context (descriptions, images, and videos) to confirm or correct automated findings. This human review acts as a safeguard for automated detection, ensuring that subsequent enforcement actions are both accurate and contextually appropriate. We also act on listings reported by rightsholders via the IP Portal, Brand Registry Portal and by users via the ¡°Report this item¡± function, which includes an option for reporting a ¡°counterfeit item¡±. Enforcement actions on confirmed infringing listings include delisting products and/or restricting stores of infringing traders. Our IP team also maintains an IP Database, updated through rightsholder partnerships, legal actions, proactive monitoring, and successful "Report this item" complaints. Where a rightsholder has previously engaged with us, we seek their direct input before taking enforcement action, further ensuring that automated detection is verified and that decisions reflect accurate IP ownership. 5. Ad-specific content moderation The review by content moderators provides an essential layer of oversight for the advertising programme. High-risk product categories or complex cases, such as those involving borderline sexualized imagery, are subjected to manual review by trained moderators. These moderators operate under detailed internal policies and receive industry-specific training to ensure consistent, accurate, and policy compliant decision making. Traders on Temu who are adversely impacted by moderation decisions within this programme may contest such decisions by submitting appeals via our internal complaint-handling system.

Temu¡¯s automated systems are designed to uphold a reliable, safe, and trustworthy environment for our users by ensuring that content, products, and interactions on the platform remain lawful, compliant with our terms and policies and are aligned with our high standards for a safe, trustworthy and compliant platform. The key objective of these automated measures is to mitigate risks identified in our risk assessment under Article 34 of the DSA, including risks related to product compliance, intellectual property rights, content integrity, the protection of minors, the safeguarding of fundamental rights, and in particular, consumer protection. Specifically, automated systems are deployed for the following purposes and to mitigate the abovementioned risks: a) Trader verification and due diligence: Reviewing all uploaded trader information and materials to verify the completeness and compliance of documentation before traders are permitted to list products. b) Large-scale listing moderation: Performing continuous, large-scale screening of listings data to ensure accuracy and compliance with relevant internal and regulatory requirements. c) Proactive IP enforcement: Flagging and blocking product listings suspected of infringing on IP rights before they reach the consumer. d) Integrity of reviews: Screening product reviews before and after publication for potentially non-compliant content, including prohibited terms and phrases and indicators of inauthenticity. e) Specialized ad-specific content-moderation: Applying a dedicated layer of moderation for promoted listings, ensuring that advertisements meet the specific compliance standards required for promotions.

This section of the Report covers the Temu mobile app and the website as available in various EU Member States (collectively, ¡°Temu¡±) and details moderation efforts taken by Temu on its own initiative between 1 July and 31 December 2025. Temu¡¯s own-initiative moderation strategy integrates automated systems, proactive measures, and human oversight. Our approach seeks to maintain a safe shopping environment through the following measures: 1. Clear, Transparent and Comprehensive Policies Temu¡¯s own-initiative moderation efforts rely on policies designed to ensure platform integrity and safeguard users from potentially non-compliant content. Our Terms of Use and Community Guidelines establish the baseline for user behaviour, content eligibility, and product listings. Beyond these, our efforts are also driven by other domain-specific policies that define the boundaries of our proactive interventions, whether automated or manual. For example, our Product Safety and Compliance Policy sets strict guidelines to ensure listings are safe, reliable, and compliant with relevant regulations and industry standards. Similarly, our Review Guidelines ensure reviews remain trustworthy. As the first level of our own-initiative moderation is performed by automated systems, these policies enable our systems to proactively reject violating listings before they go live; delete non-compliant listings visible on the platform before they reach users; and remove violating user reviews. These policies also guide human moderators who intervene in edge cases. Our platform policies are accessible in our Transparency Center and are regularly reviewed to reflect legal changes, industry standards, and emerging risks. 2. Preventive Measures to Minimize Risks Our content moderation framework starts with proactive steps to mitigate risks before they appear on the platform. We conduct thorough trader verification and continuously refine listing procedures to ensure compliance with our standards and industry benchmarks. Additional measures include proactive screening of user content, leveraging feedback from consumers, traders, and other stakeholders to identify potential risks and enhance our moderation measures. 3. Automated Tools and Manual Review Temu utilises automated systems to address emerging risks promptly. These automated systems detect, intercept or remove violating listings, such as counterfeits, or violating user reviews. In cases where automated systems cannot reach a conclusive determination, suspicious content is flagged for further review by human moderators. These moderators work in tandem with our automated tools to assess flagged content and take appropriate actions swiftly and accurately. 4. Enforcing Accountability and Protecting Users Temu takes a firm stand in holding traders accountable for attempts at distributing non-compliant or policy violating products and ¡ª where such content is detected ¡ª takes immediate action. Actions may include removing listings, suspending or terminating accounts, withholding payments or permanently banning traders if necessary. We also apply strict standards to consumers. Actions may include removing content, restricting access to community features such as reviews, and suspending or terminating accounts. Our approach to enforcement is proactive and data-driven. We continuously monitor trends and identify recurring issues to refine our processes and policies to prevent future violations. 5. Continuous Improvement, Training and Awareness Programmes Our commitment to maintaining a safe marketplace is supported by the ongoing training and development of our content moderators and the continuous improvement of our automated systems. Moderators receive ongoing training on legal standards and best practices to ensure high-quality manual oversight. Automated systems are refined through continuous feedback loops, integrating insights from manual review and experts to improve performance. Furthermore, we provide clear guidance to both traders and consumers on the platform about their roles in maintaining a safe platform. Traders receive ongoing education on compliance requirements, including intellectual property rights and product safety. We also encourage users to report suspicious content, offering clear instructions on how to flag problematic listings and reviews. 6. Substantive Changes During this reporting period, we launched a new advertising programme to enhance the visibility of selected (pre-existing) listings across Temu¡¯s interfaces. Sellers may participate after accepting applicable advertising terms. The programme incorporates ad-specific moderation procedures that enforce Temu¡¯s Advertising Service Usage Guidelines and related policies. Proactive moderation measures within this programme constitute a substantive change in our own-initiative enforcement approach, impacting the volume of actions reported in this period.

We recognize that content moderation is a demanding task that sometimes requires evaluating potentially objectionable material to ensure the safety of users on Temu. Therefore, we have established a mentorship programme to support our content moderation team and enable them to perform their duties effectively. While the risk of exposure to disturbing material is lower on a consumer product marketplace like Temu than on other digital platforms such as social media, we prioritise the psychological well-being of our staff and offer a free psychological counselling service, which employees are encouraged to utilise.

We ensure that all human content moderators undergo comprehensive training as part of their onboarding process to maintain high standards for content moderation activities. This initial training consists of an initial classroom phase followed by two weeks of practical application. Moderators are only permitted to begin live content review once they have met specific internal benchmarks for both accuracy and review volume in our testing environment. Beyond the initial onboarding, we provide targeted training sessions to reviewers every one to two weeks to refine their moderation skills in specific areas. These targeted trainings are supplemented by individualised one-on-one training sessions, where necessary. Training is delivered by team leaders through either remote video sessions or in-person workshops, incorporating interactive elements to maximize engagement. To support ongoing learning, moderators have access to reference and reading materials, including moderation guidelines, case studies, standard operating procedures, and training slide decks. We ensure that human content moderators receive specific training based on their specialised subject matter for review, such as IP and sensitive product content, to provide optimal protection for Temu¡¯s users. For example, our dedicated IP team receives regular training to ensure a solid understanding of relevant IP risks, enabling them to review and action reports from users and rightsholders appropriately. This includes training on the criteria for determining whether IP rights have been infringed and the development of skills to apply legal standards when assessing visual, conceptual, and phonetic similarities. Our product content moderators also receive specialised training on topics such as religion and adult content. Furthermore, internal reference materials are constantly updated to reflect jurisdictional and regional peculiarities on what constitutes prohibited and controlled products, and where necessary, training is conducted to ensure that product content moderators possess the knowledge to apply these nuances consistently and at scale. We also foster a collaborative learning environment where team members share insights to improve the handling of daily moderation. For example, the IP content moderation team holds periodic sessions in which moderators share their experiences in addressing recent cases.