RTFP

NKL’s moderation governance is structured through defined roles and review stages. Flagged content is processed through a daily review pool and an internal interface that assigns cases based on moderator expertise. Advanced review is conducted by the most experienced moderators and focuses on ambiguous or context-dependent material, including regional, cultural, and linguistic nuances that automated systems may fail to detect. Enforcement measures include ghosting (temporary invisibility through de-indexing while the uploader retains account access), content validation, pending takedown prior to permanent deletion, and permanent deletion of content and associated accounts in severe cases. At the organisational level, human resources are structured by moderation stage and expressed in FTEs. A Head of the Moderation Team oversees moderation operations, establishes workflows, and addresses complex cases. The basic review function conducts initial assessments and monitors general content. Advanced review handles complex cases and identifies problematic trends and procedural improvements. A specialised team processes complaints and notices (including submissions from trusted flaggers) and executes takedown requests. Channel oversight ensures that issues are addressed within specific channels. The responsibilities of moderation team members overlap across stages; however, the governance framework is designed to ensure comprehensive oversight of moderation activities.

NKL reports ToS exposure of 0.00001% and illegal-content exposure of 0.00002%. These exposure figures are intended to provide a concise, comprehensible indication of the limited level of recipient exposure to violating and illegal material, consistent with NKL's proactive own-initiative detection, rapid temporary unavailability measures for high-priority cases, and human verification of automated flags.

NKL allocates human resources for moderation using a breakdown by moderation stage, expressed in dedicated FTEs. Responsibilities overlap across stages of review, which makes it challenging to assign precise figures to each discrete task; however, NKL maintains dedicated staffing at each stage to ensure comprehensive oversight of moderation activities.

Moderators are required to possess advanced English proficiency, strong computer skills, and the ability to handle sensitive content. NKL maintains multilingual capability across all EU official languages, with additional coverage for certain non-EU languages, and uses translation tools where direct language support is not available. Keyword databases incorporate linguistic variations across all EU languages to support precise and contextually informed moderation in diverse linguistic environments.

NKL provides tool-specific indicators of detection accuracy and estimated error rates for automated moderation systems, and delineates the functional scope of each system by specifying which components of the platform interface are subject to scanning. For Vercury, accuracy for identifying blocked content is reported at 83%. For flagged-only content, initial accuracy was 2.5%, increasing to as much as 10% since June 2024. An estimated 15-20% error rate is associated with blocked content pending human review. The system scans video content exclusively. For Google SafetyNet API, accuracy varies according to threshold settings. For video analysis, the “Google High” threshold is reported at 100% accuracy, while “Google Medium” is 1.6% accurate. For image analysis, “Google Very High” is reported at 100% accuracy, “High” at 75%, and “Google Medium” at 15%. Estimated error rates are up to 98.5% for videos and 70% for images. The system scans both video and image content. For Safer, the reported accuracy rate is 10%, with a possible error rate of 90%. This tool scans video content only. For keyword search and match, no accuracy or error-rate data are available. Its scope is limited to text-based content, specifically user comments. Across systems automated moderation tools may produce a substantial volume of false positives. These limitations underscore the necessity of rigorous human oversight to evaluate flagged material contextually and to mitigate the risk of disproportionate removal of lawful content.

NKL employs automated content monitoring tools to identify, review, and manage material that may violate the platform’s ToS or legal regulations across videos, images, and text, supporting real-time monitoring and preventative measures. In particular, Vercury is a fingerprint system relying on an internal signature database: it compares content signatures against the database, generates a match score, and, based on that score, content may be blocked or sent for review, primarily targeting illegal materials; Vercury scans videos only. Hive uses AI to analyse images and videos to identify behaviour, context, or potentially harmful activity in the scene and is employed mainly to identify ToS-violating content (e.g., alcohol, violence, firearms), with actions including blocking or flagging for review; Hive scans videos and images (pictures). Google SafetyNet API uses AI to analyse images and assigns an underage risk score based on its assessment of the content and is used to flag potentially harmful content for review; qualitative parameterisation via thresholds, distinguishing outcomes depending on Google High / Google Medium for videos and Google Very High / High / Google Medium for pictures, the tool scans videos and images. Safer maintains a fingerprint database of known content to detect and flag matches efficiently and is used to flag potentially harmful or inappropriate content for review; Safer scans videos only. For text, keyword search and match is used with blocklists and grey lists: the block list prevents recipients from using specific words or phrases (and such content is not saved), while the grey list flags content for further analysis and human review; this tool scans text only (comments).

Recognising the limitations of automated systems and the risk of false positives, NKL applies layered safeguards to ensure responsible enforcement. Content flagged by automated tools is subject to additional verification before any enforcement action is taken, with manual review mechanisms operating across all tools. All content flagged by Vercury is manually reviewed to determine whether removal is warranted. Hive outputs labelled as “problematic” are manually reviewed for final decision-making, and all content flagged by the Google SafetyNet API is manually reviewed prior to enforcement. Items flagged by Safer are likewise subjected to manual review to confirm the presence of a violation. For keyword-based tools, block-list matches trigger an error message to the user without retention of the underlying data, whereas grey-list matches are manually reviewed. NKL also continuously enhances its proprietary and third-party moderation systems, driven by moderator feedback and through frequent updates to text-moderation databases incorporating new terms, slang, and evolving language patterns.

NKL uses automated means to identify, review, and manage content that may violate the ToS or legal rules across videos, images, and text, and to support preventative, real-time monitoring and prioritisation. The primary purposes include fingerprinting video and image content through Vercury to compare signatures against an internal database and generate match scores that may lead to blocking or referral for review (primarily to target illegal materials); using AI-based analysis through Hive to identify behaviour, context, or potentially harmful activity in images and videos and to support blocking or flagging for review (notably for ToS violations such as alcohol, violence, and firearms); applying Google SafetyNet API to analyse images and assign an underage risk score and flag content for review; using Safer to match against a fingerprint database of known content and flag potential matches for review; and operating keyword search and matching for text moderation via block lists and grey lists, where block lists prevent publication and grey lists flag content for further analysis.

NKL Associates s.r.o. ("NKL") employs a proactive, multi-faceted approach to content moderation designed to detect and address illegal content or content that violates the platform’s Terms of Service (ToS). Content such as CSAM, NCII or depictions of real physical violence is treated with the utmost priority. Once identified, such content is removed and subjected to internal investigation, and relevant recipient data associated with such content is reported to local law enforcement authorities to facilitate further action. When content, particularly high-priority violations such as CSAM or NCII, is flagged, its URL is promptly de-indexed, making it inaccessible on XNXX (so-called “ghosted”), while the moderation team reviews the flagged content; the content remains inaccessible until the review process is completed. In addition, XNXX employs automated means to detect and prevent the publication of text containing flagged or blacklisted terms before it becomes visible. As part of own-initiative moderation, NKL applies restrictions affecting the availability, visibility, and accessibility of content and (where applicable) recipient accounts. These include content ghosting (unindexing) as a temporary visibility restriction, content takedown (pending deletion) as an intermediate stage allowing recipients to contest removal, and content deletion (permanent deletion) as a final removal. Account-related restrictions include account termination (temporary inaccessibility) and account deletion (permanent, removal of the account and associated data/content).

NKL recognises the challenges faced by content moderators and describes support measures aimed at safeguarding both physical and mental well-being. Physical support includes access to ergonomically designed workstations (e.g., adjustable desks, supportive chairs, and monitors to reduce eyestrain) and, in many instances, opportunities for fitness and wellness programmes such as gym, swimming, or sports facility memberships, along with personalised wellness guidance. Mental-health support includes environments promoting open dialogue about stress and emotional challenges, access to confidential consultations to mitigate burnout risks, and consideration of flexible scheduling and remote options where feasible. Ongoing dialogue between moderators and senior team members is emphasised as a mechanism for sustained support, continuous feedback, and policy adaptation to evolving needs.

NKL provides structured and ongoing training to ensure moderators are equipped to perform their responsibilities. Training is conducted by the Head of Moderation Team and experienced moderators, and onboarding follows a mentorship model in which new moderators observe experienced colleagues and gradually assume tasks under supervision. Training focuses on responsible handling of sensitive content, identification and addressing of illegal material, and familiarity with moderation processes and classification methods. Moderators receive continuous feedback and have access to support resources such as a Q&A reference table and real-time communication platforms to foster collaboration and knowledge sharing.