RTFP

SHEIN’s processing of illegal content notices and related investigations is centrally led by different Content Moderation team. These teams acts as the primary governance owner for notice handling and investigation workflows, and is responsible for triaging incoming notices and assigning accountability to the appropriate internal teams. To ensure subject-matter expertise and consistent application of legal criteria, responsibilities are allocated across dedicated workstreams based on the type of illegal content reported, including, where relevant, Product Compliance, Offensive content and Review. These teams conduct the substantive assessment and implement the relevant enforcement actions within their respective remits. Human judgment remains a core element of this governance structure. SHEIN is supported by a content review function comprising over 1,400 human moderators, who perform key contextual assessments and make the majority of moderation decisions. SHEIN relies on detection of automated systems and subsequently human review to identify and address prohibited or otherwise non-compliant listings. Automated tooling plays a supporting role and does not operate as the determinant for final enforcement outcomes. Where a listing is flagged by automated detection, it is routed for subsequent human review and validation to confirm the accuracy and appropriateness of the automated detection before the final moderation action is taken. Upon receipt of an illegal content notice, the responsible team follows a standardised sequence of notification, review, investigation, and decision-making, ensuring that outcomes are aligned with applicable legal criteria and internal policies. SHEIN also tailors training and illegal content review processes by geography to reflect local legal requirements and relevant local customs, supporting consistent and compliant decision-making across jurisdictions.

We continuously check product listings and reviews published on our platform, with a particular focus on product compliance and protection of intellectual property rights, with the support of technical tools. Our systems are designed to flag certain products for review where such products have not been spot-checked for a while or where our tools detect that a product is prohibited or restricted. The review is then conducted by our moderators who take a final decision on the content. We also maintain an Intellectual Property Infringement Element Database consisting of a repository used to log and track various intellectual property infringements. This helps us document and manage instances of confirmed intellectual property violations found in product listings and notices - including external information and reports. Comprehensive records of all intellectual property infringement incidents, verification, actions taken, and resolutions are maintained in the Intellectual Property Infringement Element Database. We have a process of regular reviews that assists with analysing and identifying repeat or recurring issues to improve our intellectual property infringement detection processes. We may also rely on feedback provided by customers, sellers, and other third parties to proactively identify elements that may negatively impact the user experience. We also refer to our Prohibited Product List and any other applicable laws and regulations, including any regional specificities, to detect products that may not be compliant in certain countries.

We use both full-time employees and third party vendors to moderate content. We compute internal moderation resources by counting the number of individuals who perform content moderation tasks within our organization during the reporting period. This figure is reported as headcount (i.e., number of persons). We compute external moderation resources primarily by counting the number of individuals provided by our outsourcing service provider who perform moderation on our behalf during the reporting period. The total number of external moderators reported therefore reflects outsourced headcount where such headcount information is available. In addition to headcount-based outsourcing, we also contract with certain external moderation providers on a per-reviewed-case basis, where the service is purchased based on the number of reviewed cases rather than the number of moderators assigned. Because these contracts are not structured around dedicated or identifiable personnel and do not reliably provide a consistent, auditable headcount, therefore it is not included in the Transparency Report.

We have various functions involved in content moderation globally. These functions collaborate closely with specialised teams and focus broadly on the following areas: - Ensuring that marketplace operations meet regulatory requirements such as data protection and DSA. - Managing customer interface functionality, payment processes, customer service, and risk control. - Overseeing seller and product management covering areas like product compliance, intellectual property protection, and the moderation of offensive content. Some teams, such as our Product Review team, primarily focus on content moderation, while others, like our customer service team, incorporate content moderation as part of their broader responsibilities. They represent a mix of roles, with the extent of their moderation activities varying; some devote most of their portfolio to content moderation, while others contribute to content moderation in a more limited capacity alongside other responsibilities. Ownership of responsibilities in this process is split between different teams across the business. Each team is assigned specific workstreams based on the type of illegal content reported. We also work with third parties who provide expertise on content moderation. Such content moderation expertise may be provided directly by external moderators or as a service with no reference to personnel headcount (e.g., by type of content, volume). The teams involved in content moderation have a broader role involving different tasks and responsibilities. They are generally university graduates, but their areas of expertise may vary. • The members of our IP and privacy teams typically have a legal background, whereas the customer service (CS) team generally have background in assisting customers. • All of them are trained in content moderation and continuously improve their skills through internal trainings. In terms of languages spoken, the qualification expected from our internal moderators includes the English B2 level (or equivalent). Certain moderators also speak other languages used in the European Union, and their language proficiency is at the appropriate level needed to carry out the responsibilities of a moderator.

Shein uses automated tools to support the teams in charge of content moderation. While automated systems play a supporting role in our content moderation to help our teams quickly and firstly identify and enable decisive action against the most obvious categories of infringing content, the decisions are made by human moderators. Our system is also designed to automatically flag products that have not been checked for a while. We use continuous automated checks at scale utilising technical tools, including image recognition, to identify prohibited or restricted products.

We do not rely solely on content moderation by automated means, but use continuous automated checks utilising technical tools like image recognition and keyword detection to flag potentially violative products. Our automated tools are trained to detect keywords, images, and content patterns associated with prohibited products and other Illegal Content. If a product is flagged as prohibited or restricted, it is promptly flagged for further human review to ensure the accuracy and appropriateness of the automated detection.

The purpose of our automated tools is mainly to provide support to the moderators in the detection of potentially illegal content, in order to speed up the content moderation process and help them deal with large quantities of content, but not to replace human review. Our limited fully automated system aims at preventing the publication of product reviews containing inappropriate words (e.g., for reviews, profanities, rude or swear words), as they generally have nothing to do with the purpose of our review section or our platform at large.

Our moderators use different tools to identify potentially non-compliant products, and inappropriate content such as pornography or intellectual property infringements. We use an image search engine that assists the Intellectual Property Team and Litigation Team, Product Compliance Team, and Product Review Team to identify and remove problematic products. The tool notably enables to detect infringements of well-known trademarks. Our systems are also designed to flag certain products for review where such products have not been checked for a while or where our tools detect that the product is prohibited or restricted. The review is then conducted by our moderators who take a final decision on the content. We also use image recognition and word detection search tools designed to search and screen products against the data published on the Safety Gate of the European Union.

Shein is a global online fashion and lifestyle retail store and marketplace designed to allow users to search for, and buy, clothing and other products. Our website includes both Shein-branded listings and third-party listings from third-party sellers. We have put in place a series of measures aimed at preventing and detecting the presence of illegal or inappropriate content on our platform, including to limit: • The risks of users encountering unsafe or prohibited physical goods listed and sold on the marketplace; • The risks of users encountering inappropriate content associated with product listings (e.g., images), and • The risks of users encountering inappropriate content in user-generated reviews and comments. .Given the nature of our eCommerce marketplace, we particularly focus on the detection and removal of products infringing intellectual property rights and detection and removal of products that may not be compliant with product safety regulations.. Unsafe and prohibited products: We have a product compliance program based on the specific regional and applicable EU and Member State laws and regulations. As part of this program, we work closely with local product safety agencies, monitoring changes and developments of product safety compliance laws and regulations. We request that all sellers comply with the applicable product safety laws and regulations and conduct spot-checks to monitor compliance. We analyse products against associated regulatory requirements and on a risk adjusted approach. Our Product Governance Team aim to equip onboarded sellers with the necessary knowledge to comply with our Marketplace policies and adhere to stringent compliance standards. Once a product is listed, our moderators continuously monitor and randomly check products to ensure applicable requirements are being met. In particular, prohibited products are delisted when identified. Intellectual property: We also use a combination of automated tools and human review to detect products infringing intellectual property rights on our platform. We have a process of regular reviews that assists with analysing and identifying repeat or recurring issues to improve our intellectual property infringement detection processes. Our automated tools generally allow us to actively detect these violations, with a final decision taken by our moderators. As a result, during the reporting period, an important part of our own-initiative content moderation activity (as shown by metrics in tab 5 and 6) have resulted in the removal of non-compliant products or products infringing intellectual property rights. We have also terminated the accounts of users engaging in fraudulent activities in violation of Shein’s terms and conditions.

We utilize automated detection capabilities (including keyword and image recognition) to continuously screen and flag suspected non-compliant products, prioritizing high-risk items for manual review. For products triggering rules such as blacklisted keywords, the system automatically alerts and requires manual review, improving efficiency and reducing missed detections and false positives. We are significantly expanding the ongoing education system for our review team, launching specialized modules through a combination of online and hands-on training. These modules focus on illegal and non-compliant products, especially high-risk categories. Simultaneously, we enhance review consistency and judgment quality through scenario-based exercises based on real-world cases. Training content is adapted to local legal requirements and cultural contexts. We have over 1400 manual reviewers and continue to increase investment and expand our review and product compliance personnel. Some teams are also undergoing restructuring to improve cross-team collaboration and case flow efficiency. Review work is led and coordinated by the content review team and distributed according to task type to specialized workflows such as product compliance and review execution/review, ensuring that it is handled by personnel with the corresponding professional capabilities. We do not set or incentivize reviewers to meet targets or restrictions such as the number of approvals/rejections, in order to avoid biased decision-making. At the same time, we use methods such as sampling inspection to assess the accuracy of manual review, and we use valid reports and missed cases as feedback signals to iterate rules, training and tools. In view of the purpose of our platform as a marketplace, the type of content that we deal with (mainly fashion products, clothes or other lifestyle items), and the limited opportunities for users to publish content (essentially through the submission of product reviews), it is very unlikely that our moderators would face disturbing, shocking or harmful images or speech. Our internal moderators may always share any difficulty or issue with their peers, supervisors, line managers, human resources representatives or through other communication channels. In terms of physical support, our internal moderators benefit from the infrastructures our of company and have the opportunity to raise any requests with the internal HR teams.

Moderators generally receive initial training and supporting materials to understand and apply our methodology and procedures in relation to content moderation. We periodically update our internal documentation to keep our moderators updated as regards legal requirements in the EU. They also participate in additional and more specific training sessions in relation to, for instance, intellectual property, product compliance, or offensive content. Specifically, for product compliance, intellectual property and offensive content, we will conduct new employee training and by topic trainings for moderators, which generally consist of online courses.