Consent version: training-consent-2026-08-07-v2
1. Your choice is optional
AI model improvement is off by default. If you turn it on, you consent to the processing described in this Notice. You can use Rick without giving this permission.
Turning AI model improvement off does not prevent STARTORK from processing conversations as necessary to provide Rick, generate responses, personalize your experience, maintain memories or preferences you request, operate the service, prevent abuse, maintain security, comply with law, or provide safety features. Model-improvement consent is separate from these core service activities.
2. What you permit
If you opt in, STARTORK may select eligible text messages created after your current opt-in, including messages you send and Rick's responses, for approved AI evaluation and model-improvement activities. These activities may include:
- evaluating model quality, reliability, and safety;
- testing and comparing model behavior;
- reviewing or annotating selected examples;
- creating approved evaluation or improvement datasets; and
- training, fine-tuning, or otherwise improving AI models used or developed by STARTORK.
Selected records may, where necessary for an approved task, be reviewed or annotated by authorized STARTORK personnel or vetted service providers subject to confidentiality, access-control, security, and contractual requirements. This permission applies only to content created after the current opt-in boundary. It does not make earlier conversations eligible.
3. Service providers
STARTORK may use service providers for authorized AI inference, hosting, security, evaluation, annotation, data processing, or model-improvement activities. Where a provider processes eligible records on our behalf, it may process those records only for the authorized services it has been engaged to provide and subject to applicable contractual and security controls.
Service providers may not use eligible STARTORK conversation records to train or improve their own general-purpose models for their independent purposes unless STARTORK separately discloses that use and obtains any permission required by applicable law. Personal data may be processed outside your country as described in our Privacy Policy and subject to applicable cross-border data-transfer safeguards.
4. What is not eligible
The improvement pipeline excludes, where applicable:
- messages created before your current opt-in;
- content associated with minors, accounts under age review, or accounts whose required age status has not been confirmed;
- safety, crisis, self-harm, or similar high-risk conversations and associated risk records;
- age disclosures, age-review records, and age appeals;
- support, billing, payment, dispute, test-account, and internal QA data;
- voice recordings, call audio, call transcripts, and other voice-call records unless STARTORK introduces a separate permission;
- deleted content, subject to technical and legal limitations;
- derived memories, relationship-profile data, communication-style profiles, and similar personalization records;
- unsupported message or attachment types;
- authentication secrets, passwords, access tokens, financial credentials, payment-card information, government identifiers, and similar security-sensitive information;
- records detected as containing direct identifiers; and
- records detected as containing sensitive personal information or special-category data designated as ineligible, including health, mental-health, sexual-life, racial or ethnic origin, religious or political beliefs, genetic or biometric identifiers, or precise geolocation.
STARTORK may update its safety and eligibility filters to address new categories of high-risk information. Automated controls reduce risk but cannot guarantee that every piece of personal or sensitive information contained in free-form conversation text will be detected.
5. Safeguards before use
A message being technically eligible does not mean it will be used. Before an eligible record can enter an approved model-improvement task, STARTORK applies controls designed to confirm the permission is active, the record follows the consent boundary, the consent version remains valid, the account and content remain eligible, excluded information has not been detected, and processing is limited to an authorized purpose and dataset.
STARTORK uses a pseudonymous subject identifier instead of a direct account identifier where reasonably practicable and separates information needed to reconnect that identifier from the improvement dataset. Pseudonymization reduces identification risk but does not necessarily make information anonymous. Access is restricted according to role and purpose and may be logged or audited.
6. Feedback and other data uses
Turning on model improvement governs only the uses described in this Notice. Product analytics, security, fraud prevention, personalization, requested memories, safety systems, support operations, and other processing necessary to operate STARTORK are governed separately by our Privacy Policy and other applicable terms.
If STARTORK wishes to use feedback, reports, support interactions, voice records, or another excluded category for model training in a way not covered by this Notice, STARTORK will provide any additional notice or choice required by law.
7. How to withdraw
You may withdraw this permission at any time through the AI model improvement setting in Settings. Withdrawal does not prevent you from continuing to use Rick. After withdrawal, STARTORK will stop selecting new records and prevent withdrawn records from entering new model-improvement tasks, and will stop, cancel, remove, or invalidate pending processing where reasonably and technically possible and required by law.
Withdrawal does not affect processing that lawfully occurred before withdrawal. Information already incorporated into a completed aggregate evaluation or model update may not always be capable of being isolated or reversed. Nothing in this Notice limits rights available under applicable law.
STARTORK may retain limited records where necessary to comply with law, protect security, resolve disputes, enforce legal rights, or demonstrate consent. Information retained solely for those purposes will not become eligible for model improvement merely because it is retained.
8. Deletion and account closure
Deleting a conversation or account may cause related records that have not completed an approved improvement process to become ineligible or be removed in accordance with STARTORK's deletion procedures and applicable law. Deletion may not automatically reverse a model update or aggregate evaluation lawfully completed before the request where the relevant contribution cannot reasonably be isolated. Applicable privacy rights are not limited by this Notice.
9. Retention
STARTORK retains identifiable or pseudonymous model-improvement records only for as long as reasonably necessary for the approved purpose, security, legal compliance, dispute resolution, or demonstrating consent. Retention periods or criteria are described in our Privacy Policy or applicable retention policies. Records no longer required will be deleted, anonymized, or otherwise disposed of in accordance with applicable requirements.
10. Changes to this Notice
STARTORK may update this Notice when its data practices, purposes, eligible categories, model-improvement activities, or processing arrangements materially change. A new consent version will not make messages from an earlier consent period eligible without a new opt-in where renewed consent is required. Materially expanded uses will not be applied retroactively merely by changing this Notice. Previous versions may be retained to document what a user agreed to and when.
11. Privacy rights and contact
Additional information about STARTORK's handling of personal data and applicable privacy rights is available in our Privacy Policy.
Privacy requests may be sent to contact@startork.ai.
Data controller/operator: STARTORK PTE. LTD.
60 PAYA LEBAR ROAD, #05-09, PAYA LEBAR SQUARE, SINGAPORE 409051