Privacy Policy

We collect information that you provide directly to us, data generated automatically during your use of our Services, and information from third-party sources such as our payment processor.

If you are located in the European Economic Area (EEA) or the United Kingdom, we rely on the following legal bases:

• Contractual Necessity: To provide the Services you requested (e.g., creating your account, processing rewards).
• Legitimate Interests: To improve our AI, secure our servers, prevent fraud, and analyze how players interact with MAVAMADE.
• Consent: For specific features like marketing communications or the collection of precise geolocation.
• Legal Obligation: To comply with tax, anti-money laundering, or law enforcement requirements.

Personality, learning, and psychological inference features may also rely on explicit consent where required. See our internal lawful-basis documentation for details.

Depending on your residency, you may exercise the following rights regarding how we process, store, and utilize your personal data:

• Right to Access: Request a digital file of your personal data. Automated self-service is provided via our Privacy Portal to download your cognitive state weights, profile metadata, and gameplay session records.
• Right to Rectification: Correct inaccurate information. You can modify display names, avatars, and learning parameters directly in your Account Settings.
• Right to Erasure: Request the permanent deletion of your account. Deletion permanently purges identity credentials, state engrams, and billing records across Cognito, Firestore, and Stripe. You can trigger deletion instantly via the Privacy Portal.
• Right to Restrict Processing & Object: Limit how we use your data. You may object to the processing of your data for behavioral NPC model fine-tuning (LoRA training) or restrict the tracking of automatic telemetry signals.
• Right to Withdraw Consent: Revoke permissions at any time. You can instantly withdraw consent for non-essential analytical cookies or marketing tracking using our cookie options.
• Complain: Lodge a complaint with the Hong Kong Privacy Commissioner for Personal Data (PCPD), your local EU supervisory authority, the UK Information Commissioner’s Office (ICO), or other relevant regulators.

We respond to verified requests within 30 days. To exercise these rights, please visit our Privacy Portal. Alternatively, you may contact support here. We may need to verify your identity.

We use your information to create and manage accounts, authenticate users, provide gameplay and AI-powered narrative experiences, process purchases, grant in-game currency and inventory, and deliver customer support.

We use profile and gameplay data (including personality and learning-related fields) to personalize journeys, adapt narrator behavior, and improve educational outcomes.

We use AI interaction data to process your requests, maintain session continuity, evaluate model performance, and refine our AI systems.

We use cookie consent records and limited device or browser signals to honor your privacy choices, maintain security, and understand regional usage patterns.

With your consent where required, we may send marketing or newsletter communications and analyze aggregated trends to improve our Services.

We may use contact form and support messages to respond to inquiries and investigate reported issues.

We do not sell your personal data. We share information only with vendors who agree to strict data protection standards.

• Cloud Infrastructure & Hosting: Amazon Web Services (AWS) serves as our primary host, providing secure databases, identity provider services (AWS Cognito), and serverless infrastructure. Google Cloud Platform (GCP) is utilized for Firestore game state synchronization, AI classification services, Workload Identity Federation, and Text-to-Speech (TTS) caching. We do not use Hetzner for any cloud infrastructure.
• Analytics & Tracking: We do not integrate, load, or transmit data to PostHog, Google Analytics, or other third-party tracking/analytics SDKs. Any gameplay event latency or page-views are tracked strictly in client-side memory or logged server-side with hashed identifiers.
• AI & Voice Partners: We may transmit prompts, inputs, and text or voice choices to verified third-party API providers including Google Gemini (Vertex AI), MiniMax, Amazon Bedrock, and Fish Audio to process natural language, narration, and synthesized speech. For creative narration, we host open-weights models (such as Qwen 2.5) on our own secure vLLM instances (via Lambda Cloud or Nebius GPU nodes) exposing an OpenAI-compatible API protocol; we do not transmit any player data to OpenAI. Our partners are strictly prohibited from using your data to train their models or for any independent purposes.
• Payment & Transactional Services: Stripe handles payment checkout, subscription billing, and Stripe Connect platform payouts. Stripe provides us with order metadata, billing contact info, billing address, and secure payment method metadata. We do not store full credit card numbers: all credentials are vaulted by Stripe.
• Email Delivery & Communications: Verified email and notification services are utilized to securely deliver account, waitlist, and transaction communications.

Legal Disclosures: We may disclose data if required by law, such as in response to a subpoena, to comply with tax or regulatory requirements, to prevent fraud, or to protect the safety of our community and the security of our services.

Gracywoods Games Limited is based in Hong Kong SAR, and MAVAMADE is based in the European Union (specific location undetermined). To provide our Services, operate our games, and coordinate our AI narrative systems, your personal data may be transferred to, stored, and processed in countries outside of your home jurisdiction (including Hong Kong SAR, the European Economic Area, the United Kingdom, and the United States) where our secure cloud infrastructure and verified third-party subprocessors operate.

We ensure that all such international transfers are protected by robust legal safeguards in full compliance with applicable data protection laws (including the EU GDPR, the UK GDPR, and the Hong Kong Personal Data (Privacy) Ordinance), including:

• Standard Contractual Clauses (SCCs) & UK IDTA: For transfers of personal data from the EEA and the UK to countries without an adequacy decision, we utilize the Standard Contractual Clauses (SCCs) approved by the European Commission, supplemented by the UK International Data Transfer Addendum (IDTA), contractually binding our processors (including AWS, Google Cloud, and Stripe) to strict European and British data protection standards.
• Adequacy Decisions: We rely on official adequacy decisions where the European Commission or the UK Government has determined a third country provides an adequate level of protection (such as transfers between the UK and the EEA).
• The Data Privacy Framework (DPF) & UK-U.S. Data Bridge: For transfers to certified organizations in the United States (including Google LLC, Amazon.com, Inc., and Stripe, Inc.), we rely on their active certification under the EU-U.S. Data Privacy Framework (DPF) and the UK Extension to the EU-U.S. DPF (the UK-U.S. Data Bridge).
• Transfer Impact Assessments (TIAs): We conduct operational Transfer Impact Assessments to ensure that the local laws of any recipient country do not undermine the safeguards provided by our contractual agreements.

We retain your personal data only for as long as necessary to fulfill the operational, legal, and compliance purposes outlined in this policy. Our data retention practices are governed by automated systems, data-lifecycle rules, and statutory requirements.

• Active Accounts: All account registration, authentication, display settings, and profile/education data are retained for the life of your active account. Upon a confirmed deletion or erasure request, this data is permanently purged or fully anonymized across all stores within 30 days, subject to statutory retention exceptions.
• AI Interaction Logs: To deliver interactive gameplay and evaluate narrative adaptation, we collect prompts, inputs, voice transcripts, and AI responses. These logs are subject to strict short-term operational retention, including a 14-day automatic Time-To-Live (TTL) for active gameplay sessions and a 90-day automatic TTL for legacy Yunbora AVA conversation records. After 90 days of storage, all AI interaction logs are either permanently deleted or fully anonymized and de-identified, so they can no longer be associated with any individual user identity.
• Financial Records: Billing metadata, purchase confirmations, order histories, and Stripe platform transactions are retained for up to 7 years to satisfy audit, tax, and statutory financial obligations. Upon a confirmed erasure request, any personal identifiers within these financial records are anonymized where possible under local laws, while numeric transactional values are held securely.
• Marketing Data: Information collected for waitlist signups and marketing communications is retained for up to 24 months, or until you opt-out, unsubscribe, or withdraw your consent.
• Other Operational Records: Cookie consent logs are retained for 24 months before automatic purging; secure data export download links are deleted after 7 days; and DSAR/erasure compliance audit logs are retained for 3 years to satisfy regulatory record-keeping duties.

Our Services utilize advanced Artificial Intelligence, machine learning models, and narrative orchestration systems to provide a responsive, custom-tailored gameplay experience. Inputs, prompts, text entries, and voice choices made across our platforms are routed and processed in real-time through our dedicated middleware (known as the Brain Bridge and the Living Narrator services).

To provide a responsive experience, your inputs into the AVA AI system (such as the legacy Yunbora AVA chat) or our in-game narrator are processed on-the-fly to generate context-specific narrative progression. To execute this, our systems securely transmit data to verified third-party AI and voice processing API providers (including Google Gemini (Vertex AI), MiniMax, Amazon Bedrock, and Fish Audio). For Text-to-Speech (TTS) audio, outputs are cached in Google Cloud Storage (GCS) buckets, keyed by content hash, to ensure sub-second response times and optimize API resource efficiency.

To continuously refine NPC behavioral models, storytelling accuracy, and educational quality, we may utilize de-identified, aggregated, and keyless interaction data. This includes prompt structures, choice pathways, system latency, MBTI classification outputs, and rating feedback signals. Periodically, we compile these de-identified interactions into secure training datasets (such as golden-path files like train.assess_math.v3.jsonl). These datasets are used to train low-rank adapters (LoRAs) and fine-tune large language models (such as Llama 3 8B or 70B models) on secure, high-performance GPU compute clusters (such as Nebius H100 or Lambda Labs A10 instances). This "LoRA Forge" pipeline teaches our NPC characters and game narrators to understand our specific pedagogical, psychological, and storytelling rules without associating training inputs with any individual user identity.

While our AI personalizes gameplay, these systems do not produce legal effects or significantly affect your physical or financial circumstances. Personalized gameplay is driven by a specialized Trait Weighting System (TWS) and Key Response Parameters (KRP) framework that dynamically maps personality characteristics and cognitive progress. We monitor a real-time state index representing Performance, Memory, Emotion, and Stress (P/M/E/S) weights derived from player actions. Educational scaffolding, pacing, cognitive operations, and representation ladders are adapted to fit your unique learning profile. Standard manual review and support pathways are always available to handle any profile-level adjustments.

Our educational and game experiences are built upon rigorous pedagogical, psychological, and algorithmic frameworks designed to offer deep, adaptive, and personalized learning journeys:

• Pedagogical Scaffolding & Dynamic Alignment: We design learning pathways that adaptively scaffold concepts based on cognitive progress. Our system measures active responses to introduce intermediate supports or higher-order challenges, matching the optimal cognitive zone of each learner.
• Real-Time Adaptive Personalization: We employ a sophisticated Trait Weighting System (TWS) and monitor Key Response Parameters (KRP) to adjust storytelling, difficulty levels, and NPC conversational styles. This adaptation is informed by real-time Performance, Memory, Emotion, and Stress (P/M/E/S) state scoring computed locally or dynamically serverless to respond to player engagement.
• Ethical AI Guardrails & Player Agency: We adhere to strict ethical frameworks ensuring player agency is respected at all times. Our AI narrators and autonomous agents operate within tight behavioral guardrails to prevent manipulative or deceptive interactions, and never force player choices or simulate coercive emotional responses.

We are committed to green computing and minimizing the environmental and computational carbon footprint associated with running complex large language models and modern 3D graphics:

• Resource-Optimized Compute: Our backend architecture is designed around serverless and microservice paradigms (including AWS Lambda and Google Cloud Run) that automatically scale down to zero when idle, avoiding the energy waste of continuously running inactive servers.
• Dynamic Session Context Compression: To reduce the token counts and heavy energy-intensive CPU/GPU cycles required by Large Language Models, we implement dynamic session compression. We summarize historical conversational elements and utilize high-efficiency Firestore vector database embeddings, maintaining a highly condensed and effective LLM context window.
• Carbon-Efficient Cloud Placement: We host our primary cloud database, compute nodes, and AI pipeline resources with cloud providers (AWS and Google Cloud Platform) in regions selected for their high reliance on carbon-neutral and renewable energy grids.
• Asset Caching & CDN Optimization: Heavy 3D assets, textures, and generated Text-to-Speech (TTS) audio clips are cached aggressively on global Content Delivery Networks (CDNs) to prevent redundant rendering and generation pipelines, minimizing network travel distance and overall cloud computing power.

Mavamade.com uses cookies and similar technologies for essential functions (such as keeping you signed in across related sites), remembering your cookie consent choices, and supporting checkout recovery.

When you set cookie preferences, we store your choices locally and sync them to our servers. If you are signed in, your consent record may be linked to your account so preferences stay aligned across sessions.

Essential cookies include authentication tokens shared across *.mavamade.com subdomains. Analytics and marketing: Mavamade-fe has cookie consent categories for analytics and marketing but does not currently load Google Analytics or similar third-party analytics SDKs.

LitsGO uses a signed Cloud-CDN-Cookie (Cloud-CDN-Cookie) to deliver authenticated 3D assets. LitsGO has no analytics SDK; some session page-view or action counts exist in client memory only and are not sent to any servers unless explicitly specified. LitsGO does not currently display a separate cookie-consent banner; gameplay authentication and asset cookies are necessary for the game to function.

You can manage or delete cookies through your browser settings. Disabling essential cookies may prevent sign-in, checkout, or game asset loading from working correctly.

GracyWoods Games Limited implements robust technical and organizational security measures to protect your information against unauthorized access, alteration, disclosure, or destruction, ensuring the integrity and confidentiality of your personal data:

• Encryption: All data in transit is protected via TLS 1.2+ encryption using secure HTTPS and WSS protocols on Amazon CloudFront and API Gateway endpoints. Passwords and identity credentials are managed securely through AWS Cognito user pools; we never receive or store plain-text passwords. Session tokens are protected via strict SameSite policies, secure flags, and short idle timeouts, while premium 3D game assets are delivered securely using signed Cloud-CDN-Cookie configurations.
• Anonymization & De-identification: We strip personal identifiers (such as user IDs, names, and contact information) from player inputs, prompts, and dialogue logs before compiling de-identified "golden path" datasets (such as train.assess_math.v3.jsonl) for offline fine-tuning. Our "LoRA Forge" pipeline trains low-rank adapters (LoRAs) and fine-tunes Large Language Models (such as Llama 3 8B and 70B models) on secure GPU clusters (Nebius H100 and Lambda Labs A10 instances) using only de-identified cognitive state weights and anonymized conversation engrams.
• Access Controls: Database and serverless backend access is restricted on a strict "need-to-know" basis. Our microservices utilize keyless authentication via Google Cloud Platform (GCP) Workload Identity Federation (WIF) instead of static service account credentials. In addition to AWS IAM role permission boundaries and Cognito access policies, Firestore database endpoints are protected using robust collection-level and cell-level security rules that validate authenticated user tokens (ensuring players can only access their own data).
• Multi-Factor Authentication (MFA/TOTP) and Dynamic Risk-Based Access Control: To further protect user identity for Personal (B2C) accounts, we support a second sign-in step using a free authenticator app (such as Google Authenticator, 1Password, Authy, or Bitwarden) which generates Time-based One-Time Passwords (TOTP). To optimize user experience while maintaining robust protection, we employ dynamic risk-based MFA rules: personal accounts are exempt from reauthentication challenges if they have been logged out for less than 2 hours, provided that no risk triggers are present. A full TOTP/MFA reauthentication is strictly mandatory if a sign-in is attempted from a location (detected via request headers) other than the country or region where the account was set up, or if spam/risk behavior is detected. If you do not utilize an authenticator app, secure email verification serves as your mandatory MFA fallback, requiring a secure 6-digit code delivered to your registered email address using Amazon Simple Email Service (SES).

We respect global data privacy standards and implement robust safeguards aligned with GDPR (EU/UK), CCPA/CPRA (California), and PDPO (Hong Kong) compliance frameworks:

• Data Protection Impact Assessments (DPIA) & Prior Consultation: Because our AI frameworks perform systematic and extensive evaluations of personal aspects via automated processing (including cognitive profiling and trait weighting), we proactively conduct Data Protection Impact Assessments (DPIAs) to ensure privacy by design. Our DPIAs contain a systematic description of processing operations, assess the necessity and proportionality of these operations, evaluate risks to the rights and freedoms of our players, and outline strict safeguards and security measures to mitigate such risks. We review these assessments whenever there is a change in processing risk. Should a DPIA indicate high residual risks that cannot be appropriately mitigated, we will proceed to prior consultation with the relevant supervisory authority pursuant to GDPR Article 40.
• Data Minimization and Purpose Limitation: We collect and process only the minimal set of personal information required to run, personalize, and secure our AI educational games. We never repurpose user data for unrelated commercial activities.
• State-of-the-Art Technical Safeguards: All user personal information, game logs, and cognitive profiles are stored securely. We enforce AES-256 encryption for data at rest, TLS 1.2+ transit encryption for all APIs, and utilize keyless Workload Identity Federation (WIF) instead of static service account credentials.
• Direct Self-Service Integration: We believe in user-empowered privacy. Our custom, automated Privacy Portal allows you to instantly trigger self-service data portability exports (generating direct, machine-readable JSON files) or request complete, permanent account and data erasure across Cognito, Firestore, and Stripe without waiting for manual support approvals.

Our Services integrate with third-party platforms as described in the Sharing section, including Stripe for payments, Google services for fonts, Drive import (when you choose to connect), Firestore workflow data, and AI generation, and CDN providers for 3D assets and media.

When you complete a purchase, you may interact directly with Stripe-hosted checkout or billing pages governed by Stripe’s privacy policy.

Waitlist submissions on some pages may also be submitted to Google Forms in addition to our own contact API.

Our websites and games may contain links to external sites (social media, partner merchants, and documentation). We are not responsible for the privacy practices of those third-party sites. Please review their policies when you leave our Services.

GracyWoods Games Limited is deeply committed to protecting the privacy of young players, and our Services are designed for general audiences who are at least 13 years of age (or the minimum age of digital consent in your local jurisdiction).

Under-13 Restriction and Age-Gating: Our websites, dashboard interfaces, and gameplay systems are not intended for, nor directed to, children under the age of 13. We implement age-gating mechanisms, such as date of birth (DOB) inputs during registration, to prevent individuals under 13 from establishing accounts or submitting personal data.

Discovery and Account Termination: We do not knowingly collect, store, or process personal data from children under the age of 13. If you are a child under 13, please do not attempt to register an account or transmit any personal information to us. In the event that we discover or are notified that we have inadvertently collected personal data from a child under 13, we will immediately suspend or terminate the account and execute an automated, permanent purge of all associated records across our database registries.

System-Wide Deletion (Privacy-Handler SLA): When an account is flagged for under-age cleanup, our unified privacy-handler erasure orchestrator (erasure_service.py) runs a system-wide purge. Within our strict compliance SLA, we permanently delete the user's Cognito identity profile, wipe all adaptive state engrams and dialogue logs from Firestore, delete active game session cache, and scrub transactional metadata from our Stripe billing systems.

Education and Institutional Whitelisting (Bypass Provision): We whitelist our products and services for authorized business and educational organizations. Under these arrangements, partnering schools and institutes may willingly sign up and provision accounts for children under the age of 13. For these users, GracyWoods Games Limited processes player data strictly in accordance with local regulations, COPPA, and GDPR, acting as a secure data processor under the master subscription of that partner institution. Control, management, and ultimate ownership of this data remain with the respective educational institute or business partner.

Parental and Guardian Rights: We encourage parents and legal guardians to monitor their children's online activities. If you are a parent or guardian and believe that your child under 13 has registered an account or provided us with personal data, you have the right to request access, correction, or permanent deletion of their data at any time. To exercise these rights, please contact our support team here or submit a request directly via our automated Privacy Portal.

Minors and Age of Majority: For users who are over 13 but under the legal age of majority in their jurisdiction (e.g., under 18), we restrict access to certain interactive commerce, community discussion channels, and CREME marketplace transactions to prevent unauthorized commitments, while processing their gameplay choices with enhanced privacy filters.

This section applies solely to residents of the State of California ("consumers" or "you") and supplements the other provisions of this Privacy Policy. GracyWoods Games Limited implements these disclosures to comply with the California Consumer Privacy Act of 2018 (CCPA), as amended by the California Privacy Rights Act of 2020 (CPRA).

Categories of Personal Information Collected and Disclosed: In the preceding 12 months, we have collected, and may disclose for business operational purposes to our secure cloud and API subprocessors, the following categories of personal data (defined under the CCPA/CPRA):

• Identifiers: Such as your Cognito preferred username, email address, custom player database identifiers, short-lived session authorization codes bound to IP addresses (cleared immediately after auth flow), and secure billing identifiers from Stripe (e.g., Stripe Customer IDs, Connect account IDs, or checkout session tokens).
• Commercial Information: Such as in-game wallet balances (gems and coins), transactional history, items purchased, virtual inventory configurations, Stripe customer identifiers, and subscription/billing statuses. (We do not collect or store full payment card numbers: payment processing is handled directly by Stripe).
• Internet, Game, or Other Electronic Network Activity Information: Such as game event timestamps, routing and per-turn latency logs, session phases, narrative and dialog choice history, and signed Cloud-CDN cookies (Cloud-CDN-Cookie) used to verify access to premium 3D assets. On-device local storage is used exclusively to store UI settings (such as language, theme, audio preferences, and dashboard layout) and local client metrics (GPU renderer type, deviceMemory, network hints) to dynamically auto-tune graphics performance.
• Sensitive Personal Information: Such as inferred psychological and cognitive profiling attributes, self-reflection answers, learning styles, personality trait weights (Trait Weighting System), and Performance-Memory-Emotion-Stress (P/M/E/S) state scores used dynamically to personalize NPC interactions and adaptive narrative trajectories. Plain-text passwords are never collected or stored; they are handled via AWS Cognito identity user pool hashing.
• Geolocation Data: Rough country boundaries derived serverless at the edge from Amazon CloudFront request headers (cloudfront-viewer-country). We do not collect or track precise physical coordinates (GPS).

No Sale or Sharing of Personal Information: GracyWoods Games Limited has not, does not, and will not "sell" your personal information, nor do we "share" your personal information for cross-context behavioral advertising (as defined under the CCPA/CPRA). We do not load third-party marketing trackers, Google Analytics, or other behavioral advertising pixels across our web or game clients. Furthermore, any data transmitted to our verified AI and voice partners (including Google Gemini, MiniMax, Amazon Bedrock, and Fish Audio APIs) is bound by strict processor agreements that prohibit these partners from using player data for their own independent purposes, model training, or commercial distribution.

Right to Limit the Use of Sensitive Personal Information: You have the right to limit the use or disclosure of your Sensitive Personal Information (such as personality trait logs, cognitive/learning styles, or self-reflection dialogs) to only those uses which are necessary to perform the services reasonably expected by an average consumer. Because we utilize these cognitive indicators strictly to compute real-time narrative adaptation (such as personalizing story branches, Trait Weighting, and in-game learning pacing), we do not use this data for any secondary profiling or non-essential commercial purposes. You can request to limit, opt-out of offline fine-tuning (our "LoRA Forge" pipeline), or completely erase your profile state at any time via our automated Privacy Portal.

Your Other CCPA/CPRA Rights: California residents have the Right to Know/Access their collected data (available for instant download via our Privacy Portal), the Right to Delete their account (purging Cognito, Firestore, and Stripe registries via our Privacy Portal), the Right to Correct inaccurate info via Account Settings, and the Right to Non-Discrimination.

To exercise your rights, please visit our Privacy Portal or contact us here (with the subject line "California CCPA Request"). We will verify your identity via multi-factor authentication or account login verification before processing any request.

This Refund Policy governs the billing, purchases, and virtual token transactions executed across MAVAMADE, Yunbora, and LitsGO services:

• Finality of Digital Sales: All purchases of digital goods, virtual currency (such as gems, coins, or credits), unlocked character cosmetics, in-game gear, and activated premium subscription tiers are final and non-refundable once the digital content is made available, consumed, or activated.
• EU and UK Statutory Right of Withdrawal: If you are a resident of the European Union or the United Kingdom, you have a statutory right to withdraw from digital purchases within 14 days without giving any reason, provided that you have not explicitly consented to the immediate performance/consumption of the digital asset (e.g., spending the gems, downloading the asset, or logging into the premium subscription) and acknowledged the consequent forfeiture of your withdrawal right.
• Technical Transaction Failures: In the event of a verified transactional error, server-side delivery discrepancy, or billing failure where an asset or subscription tier is charged but not successfully delivered, you may submit a review request to our support team within 30 days of the transaction. Verified errors will be resolved via account asset correction, credit, or full refund to the original payment method.
• CREME Campaign Pledges (Card Authorization Holds): When you pledge toward a creator production campaign on the marketplace, a Stripe SetupIntent is initiated to securely authorize and vault your card details. No charges are captured immediately. If the campaign successfully reaches its published pledge goal and production is triggered, your vaulted payment method will be charged the total pledge amount in USD. If the campaign does not reach its pledge goal, the authorization is canceled or expired, and no charge is made.

For any billing-related inquiries, please visit our Privacy Portal or contact us directly.

We may update this Privacy Policy from time to time to reflect changes in our practices, Services, or legal obligations.

When we make material changes, we will post the updated policy on this page and update the effective date. Continued use of our Services after changes take effect constitutes acceptance of the revised policy.