TemboAI is operated by The Sailscasts Company LTD, doing business as TemboAI. We operate an African language-data network and application products such as TemboAI Chat. Kelvin is the accountable privacy owner; contact [email protected]. This notice explains what we collect, why we use it, and the choices available to you.
1. Information we collect
Account and access information
When you use email or Google sign-in, we receive information such as your name, email address, profile image, authentication records, and account security events.
Contributor applications and mission work
If you apply to contribute, we collect the language and task qualifications you provide, where and how you use the language, review notes, and qualification decisions.
For mission work, we may collect:
- Voice recordings, text, transcripts, translations, evaluations, and cultural explanations you submit.
- The mission, prompt, language, locale, recording duration, capture method, timestamps, and quality-check results connected to the submission.
- Consent version, accepted purposes, licence, retention period, and a receipt showing when consent was given.
- Review decisions, reasons, appeals, and appeal outcomes.
- Technical and anti-abuse signals needed to protect contributors, patrons, customers, and mission funds.
Every mission displays its intended uses, prohibited uses, compensation, review standard, and consent terms before contribution.
Patron and payment information
We record Cowrie purchases, mission allocations, payment references, fees, and ledger entries. Bachs processes checkout, connected-account onboarding, bank verification, and payouts. TemboAI does not store a contributor's full bank account number after registration; we retain provider identifiers, a cryptographic fingerprint, bank name, account name, and masked account number where needed to operate and reconcile payouts.
TemboAI Chat and application data
When TemboAI Chat is available, we process prompts, responses, feedback, and application settings needed to provide it. Chat content is not added to a training dataset unless a separate, explicit consent flow says what will be used and for what purpose.
Optional conversational mission follow-ups
Some conversation missions may offer an independently unticked choice to have an AI character react to your confirmed words. If you choose it, continuing an answer sends the confirmed transcripts and the mission questions to the processor named in that choice to generate and check a bounded follow-up. This happens before final submission, including when a mission is later abandoned. Recordings, account identifiers and payment details are not included in this feature's model requests; please still keep private information out of your words.
The in-product notice must identify the actual processor and explain its retention, processing location, international transfers and restrictions on provider training before this feature is enabled. Those arrangements require owner review; this draft does not claim zero provider retention or that data remains in Nigeria. TemboAI's initial adapter is configured for OpenAI-compatible text inference, but configuration alone is not approval for collection or transfer.
You may decline and use prepared questions at the same disclosed reward and review criteria. Switching off stops new follow-up requests; it cannot recall text already processed. This choice does not grant model-training, evaluation, dataset-sale, publication or voice-cloning rights. Those uses remain separately controlled. We record the exact notice, your affirmative request, the generated question, model and policy versions, and a digest binding it to your earlier answer. The short-lived generation cache is cleared after task expiry by the retention job; submitted question/answer provenance follows the submission's approved retention and suppression rules. Minimal consent/security audit records follow the separately disclosed audit schedule. Contact support using the mission or submission reference for access or deletion requests, including abandoned missions.
Device and usage information
We may collect browser and device information, IP-derived security signals, pages used, timestamps, diagnostics, and similar operational data.
2. How we use information
We use information to:
- Authenticate users and operate accounts.
- Receive, quarantine, review, and quality-check mission submissions.
- Verify qualification, enforce mission limits, detect duplicate or abusive activity, and investigate fraud.
- Create consent and provenance records and prepare rights-defined datasets.
- Pay contributors, reconcile patron funding, and maintain an auditable Cowrie ledger.
- Train and evaluate internal language-AI models only under the exact affirmative permission recorded for your contribution. New required-use missions explain this purpose before you join; older optional choices retain their original meaning. Evaluation-only permission does not permit training.
- Deliver or license a dataset only under independently approved, specifically explained third-party data permission; internal model training permission does not include that activity.
- Operate TemboAI products, communicate service notices, and comply with law.
3. Model use and dataset licensing are separate
Base collection consent alone does not permit model training or automated evaluation. Under the Contributor Agreement, new required-use missions explicitly ask for collection, human review and internal language-AI training/evaluation before participation. The confirmation starts unchecked. You may decline and not join, or withdraw later.
Older missions that offered optional internal evaluation and training retain those independent choices, exact descriptions and versions. Declining them does not retrospectively reduce an earned reward or convert a collection-only receipt into model permission.
Internal commercial model training means TemboAI develops and operates models in its own services, including paid products under the disclosed permission. Third-party dataset licensing means another organisation receives data or a licence for its own use. The latter is excluded from the current Contributor Agreement and internal model-use choices. Voice cloning, speaker identification, public recordings/transcripts and downloadable weights trained using contributions are also excluded. Additional uses need separate, specific approval and affirmative consent.
A voice recording can identify someone even when account details are removed. Do not assume that calling it a language dataset makes it anonymous or stops it being personal data.
4. Who receives information
Our service-provider categories are application hosting and infrastructure, file storage, and transactional email delivery. Cloudflare R2 stores uploaded files. Account and mission records are handled by our existing application database and cache; email services process the contact information and message content needed for sign-in and service notices. These services support account operation, contribution review, rewards, security and the uses you have consented to. Processing may occur outside Nigeria. We remain responsible for meeting applicable data-protection and international-transfer requirements; accepting the Contributor Agreement does not waive your rights or provide blanket consent to additional uses.
Other authorised recipients include:
- Bachs for checkout, connected-account onboarding and payouts where you use those services.
- Qualified reviewers and operators who need access to decide work or an appeal.
- Third-party dataset customers only if a separate approved permission and matching contract authorise that specific delivery; the current internal model-use choices do not.
- Authorities or other parties when required by law or necessary to protect rights, safety, and platform integrity.
5. Storage and security
Uploaded mission files are stored using Cloudflare R2. Access through TemboAI is restricted by account, mission and review permissions. Structured records are stored in our database; sensitive provider identifiers use application-level encryption. We use access controls, signed webhook verification, immutable ledger entries, checksums, quarantine states and operational monitoring appropriate to the risk. These controls do not make a voice recording anonymous.
No system is perfectly secure. Please do not include real names, phone numbers, account details, addresses, or another person's private information in a mission unless the mission expressly requires and permits it.
6. Retention, suppression, and deletion
Account records are retained while your account is active and as needed for security, tax, payment, dispute, and legal obligations.
For contributions under the new Contributor Agreement, original recordings and their associated submitted text are retained for no more than three years from submission, with earlier deletion when no longer needed for permitted purposes or required by a valid request. Older receipts remain tied to their own recorded terms and retention deadlines. Rejected or abandoned material is not available for training/evaluation and must not be kept merely because it might become useful.
Consent, security, payment and dispute records are separate from source recordings. They may be kept only where justified by a separate lawful basis, with restricted access. Backup restoration must not return withdrawn or deleted sources to active use. Contact us about a specific record or retention exception; we will explain its basis and applicable retention or review criteria.
Withdrawal blocks new dataset builds, training and evaluations covered by the withdrawn permission, and starts the applicable source-data deletion process. It does not promise to erase statistical influence from a model already trained or deployed, recall model copies already lawfully released, or automatically unlearn a contribution. Public data/model-weight release is not authorised by the current consent choices; any future proposal requires separate terms explaining the practical recall limits before you choose. Backup and legally required records may remain only under the disclosed, justified retention and access restrictions.
7. Your choices and rights
Depending on applicable law, you may ask to access, correct, export, delete, or restrict personal information; object to certain processing; withdraw consent for future processing; or request suppression from future dataset releases. Withdrawal does not invalidate processing that was lawful before withdrawal.
Contact [email protected]. We may need to verify your identity before acting.
8. Age requirements
Language-data collection, including voluntary and paid contributor missions, is for people aged 18 or older. TemboAI Chat and other applications are not intended for children under 13, and local age requirements may be higher. We do not knowingly accept paid mission work from minors.
9. International transfers
Processing may occur outside Nigeria. We remain responsible for meeting applicable data-protection and international-transfer requirements; accepting the Contributor Agreement does not waive your rights or provide blanket consent to additional uses.
10. Changes and contact
We may update this policy as the platform changes. Material changes will be communicated through the product, website, or email as appropriate.
Questions or requests can be sent to [email protected].