Terms of Service
Effective date: [EFFECTIVE DATE] · Last updated: [LAST UPDATED]
1. Who these terms are between
These Terms of Service (the “Terms”) are an agreement between Numen Labs LLC, a limited liability company registered in [JURISDICTION OF INCORPORATION] with registered address [REGISTERED ADDRESS](“we”, “us”, the “Operator”), and you, the person or organisation using DATAtokeniz (the “Service”). If you use the Service on behalf of an organisation, you confirm you are authorised to bind it.
By using the Service you accept these Terms. If you do not accept them, do not use the Service.
2. What the service does
DATAtokeniz converts material you supply — pasted text, uploaded documents, images, audio and video files, or a public URL it fetches for you — into training-ready datasets (JSONL and related formats) with a per-record provenance trail, an automated quality report, and an EU AI Act Article 10 data-governance record. When you are signed in you can save a finished dataset to your own Google Drive.
- The lane the interface uses is synchronous. When you paste text or upload a file in the app, your material is read, processed and streamed back within the single request that carried it, and no part of it is written to storage we own.
- There is a second, durable lane, and it does retain your content while a run is in flight. A run started at
POST /api/jobsprocesses a file you reference in your own Google Drive. It is built on a durable workflow runtime in which each stage is a step whose result is written down, so that a run interrupted by a crash, a retry or a platform timeout resumes instead of starting over. That persistence is the mechanism, not a side effect: the source bytes fetched from your Drive and the compiled dataset are each step results, so each is held by that runtime for as long as it keeps the run record. Section 6 and the Privacy Policy set out what that means. This lane is reached through the API; the interface does not currently start one. - Without an account, we create no account for you and store neither your source material nor your dataset. One thing is written: a counter used to enforce the demo rate limit, keyed by a salted, keyed digest of your IP address rather than the address itself.
- Signed in, we record a reference to the run — the number of sources, the profile used, the quality score, billable counts and cryptographic hashes — never the dataset bytes. Your outputs go to your own Google Drive or your own browser download.
- Some steps depend on configuration. Model-assisted enrichment, the model-backed guide, image analysis and audio/video transcription only run when the operator of the deployment has configured the corresponding provider keys. Where a capability is unavailable, the Service says so and continues deterministically rather than silently substituting something else.
- Records are labelled by origin. Content that was extracted from your source and content that was generated by a model are marked as such and are never mixed silently.
Plan limits (input size, runs per month, pages per month) are published on the pricing page and enforced in the Service.
3. Accounts
Accounts are created by signing in with Google. You are responsible for the security of the Google account you use and for all activity under your DATAtokeniz account. Each account is owned by exactly one user; there are no shared workspaces or roles. Tell us promptly at [SECURITY CONTACT EMAIL] if you believe your account has been compromised.
4. Your material and the rights you warrant
“Input” means anything you supply to the Service. “Output” means the datasets, reports and records the Service produces from it.
Before each run the Service asks you to confirm two things. Both are warranties under these Terms:
- that you own the Input or otherwise hold all rights, licences and permissions necessary to process it as you have instructed, including any permission required from the people the material is about; and
- that the licence you declare for the Input is accurate. That declaration is written into the dataset and its Article 10 record, so it becomes part of the evidence you rely on later.
Changing the staged sources invalidates a previous attestation and you will be asked again. You are solely responsible for the lawfulness of your Input and for any consequence of an inaccurate rights or licence declaration.
If your Input contains personal data, you are the controller of that data and we process it only on your instructions. See the Privacy Policy, and put a data processing agreement in place with us before processing personal data at scale: [DPA CONTACT OR LINK].
5. Acceptable use
You must not use the Service to:
- process material you do not have the right to process, including material subject to a licence or terms that forbid text and data mining or dataset creation;
- process material whose creation, possession or distribution is unlawful where you or we operate;
- build datasets intended to identify, profile, track or discriminate against people unlawfully;
- submit URLs in order to attack, scan, overload or gain unauthorised access to any system, including our own infrastructure;
- upload malware, or content designed to manipulate the Service’s model-assisted steps into behaviour outside their documented purpose;
- circumvent plan limits, rate limits or authentication, or share an account to exceed a plan’s entitlement;
- resell, sublicense or provide the Service to third parties as your own service without our written agreement;
- misrepresent an Article 10 record, quality report or provenance trail — for example by presenting a preview record as a countersigned one, or by altering a record while continuing to present it as issued by the Service.
We may apply automated limits and technical protections (rate limiting, size caps, network restrictions on URL fetching) and may refuse a request that trips them.
6. Intellectual property
You keep everything you bring and everything you get out. You retain all right, title and interest in your Input and your Output. We claim no ownership of either, and we take no licence to use, publish, disclose or otherwise exploit your Input or Output for any purpose other than performing the processing you asked for in the request that carried it.
- We do not use your Input or Output to train, fine-tune or evaluate any model.
- In the synchronous lane we do not retain your Input or Output after the request that processed it.
- In the durable lane we do, for the life of the run record.The workflow runtime persists each step’s result so a run can resume, and two of those results contain your content: the bytes fetched from your Drive, and the compiled dataset. The stage-event stream for a run is also retained so a reconnecting client can replay it. Retention there is governed by the workflow runtime we build on, not by application code we control, and we will not state a period we cannot verify: [DURABLE RUN RETENTION PERIOD — CONFIRM WITH THE WORKFLOW RUNTIME PROVIDER AND STATE IT HERE]. While it is held, it is held only to complete the run you started; the commitments above about training, sale and disclosure apply to it unchanged.
- When model-assisted steps are enabled on the deployment you use, the relevant excerpts are transmitted to the configured model provider solely to produce your Output. Those providers handle it under their own terms; the sub-processors are listed in the Privacy Policy.
- The guide that recommends an output profile receives only source names, types and sizes — never the contents of your sources.
We retain all rights in the Service itself: the software, the canonical schema, the interface, the documentation and the DATAtokeniz name and marks. Nothing in these Terms transfers them to you. Any feedback you send us may be used without obligation, but feedback is not Input and you should not include confidential material in it.
7. Plans, billing, cancellation and refunds
- Free. No card, no charge. Subject to the published caps, which we may adjust with notice.
- Pro. A recurring subscription at the price shown on the pricing pageat the time you subscribe, plus any usage that exceeds your plan’s included pages, charged at the published overage rate. Prices are exclusive of taxes unless stated; you are responsible for any applicable VAT, GST or sales tax unless we are required to collect it.
- Payments. Payments are processed by Stripe. We do not receive or store your card details. Your subscription renews automatically each billing period until cancelled.
- Cancellation. You can cancel at any time from the billing portal reachable in the app. Cancellation takes effect at the end of the period you have already paid for; you keep Pro access until then, after which the account reverts to the Free caps. We do not delete your account when a subscription ends.
- Metering. Billable units — pages, examples, records, input bytes, media minutes and images — are measured from the finished dataset, so any invoice line can be reconstructed from the Output you received. A page counts once per source page, not once per output record.
- Refunds. [REFUND POLICY]. This does not affect any statutory right you have as a consumer, including any right of withdrawal under [CONSUMER LAW REFERENCE].
- Price changes. We may change prices with at least [NOTICE PERIOD] notice before they take effect for your next renewal.
- Non-payment. If a payment fails we may downgrade or suspend Pro entitlements after notifying you at the email on the account.
8. The service is provided as-is
To the maximum extent permitted by law, the Service is provided “as is” and “as available”, without warranties of any kind, express or implied, including any implied warranty of merchantability, fitness for a particular purpose, non-infringement, accuracy or uninterrupted availability.
In particular, and stated plainly because it matters more here than usual:
- The quality score, PII scan and secret scan are automated, pattern-based checks. They will miss things and they will flag things that are fine. Records flagged as containing personal data or credentials are excluded from exports, but exclusion is a safety net, not a guarantee that an export is free of either. Review your Output before you rely on it.
- Extraction is imperfect. Text extraction, layout handling, transcription, image analysis and language detection can all be wrong, and some formats and capabilities are unavailable without provider configuration.
- An Article 10 record is documentation, not certification. It records what the pipeline observed about your data. It is not legal advice, not a conformity assessment, and it does not make you or your AI system compliant with the EU AI Act or any other law. Whether an obligation applies to you depends on your role and risk class, and you should confirm that with your own counsel.
- Model-assisted output is probabilistic. Where a model contributes, the record is labelled as model-generated; that label is a warning to verify it, not a claim that it is correct.
We do not warrant that the Service will be available without interruption, and we may change, suspend or discontinue features. Where a feature is removed, we will make reasonable efforts to give notice to paying customers.
9. Limitation of liability
To the maximum extent permitted by law, neither party is liable for indirect, incidental, special, consequential or punitive damages, or for loss of profits, revenue, goodwill, or data, arising out of or relating to the Service, even if advised of the possibility.
To the maximum extent permitted by law, our total aggregate liability arising out of or relating to these Terms or the Service is limited to the greater of (a) the fees you paid us in the twelve months before the event giving rise to the claim, and (b) [LIABILITY FLOOR AMOUNT].
Nothing in these Terms limits liability that cannot lawfully be limited, including liability for death or personal injury caused by negligence, for fraud or fraudulent misrepresentation, or any other liability that applicable law does not allow to be excluded. If you are a consumer, your statutory rights are unaffected.
10. Indemnity
You will indemnify and hold us harmless against claims, damages and reasonable costs arising from your Input or your use of the Output — in particular any claim that processing your Input infringed a third party’s intellectual property, breached a licence or contract, or violated data-protection law. This does not apply to the extent the claim arises from our own breach of these Terms. [CONFIRM INDEMNITY SCOPE WITH COUNSEL].
11. Suspension and termination
- You may stop using the Service at any time. Signing out does not delete your account.
- To delete your account and the records associated with it, email [PRIVACY CONTACT EMAIL]. Deleting the account removes your profile, usage counters, run references and any stored Google credential. Two things it does not reach, stated plainly rather than left to be discovered: datasets you have already downloaded or saved to your own Drive, which are yours and outside our control; and any durable-lane run state still held by the workflow runtime, which expires on that runtime’s own schedule — [CONFIRM WHETHER DURABLE RUN STATE CAN BE DELETED ON REQUEST, AND SAY SO HERE].
- We may suspend or terminate access if you materially breach these Terms, if we are required to by law, or if your use threatens the security or availability of the Service. Where the law and the circumstances allow, we will give notice and an opportunity to fix the problem first.
- On termination, sections 4 (warranties), 6 (intellectual property), 9 (liability), 10 (indemnity) and 13 (governing law) survive.
12. Changes to these terms
We may update these Terms. For material changes we will give at least [NOTICE PERIOD] notice by email to the account address or an in-app notice before they take effect. Continued use after that date means you accept the updated Terms; if you do not, cancel before the effective date.
13. Governing law and disputes
These Terms are governed by the laws of [GOVERNING LAW — TODO], without regard to its conflict-of-law rules. The courts of [EXCLUSIVE VENUE — TODO] have exclusive jurisdiction over any dispute, except that either party may seek injunctive relief in any court of competent jurisdiction. If you are a consumer, you keep the protection of the mandatory law of your country of residence and may bring proceedings there. [ARBITRATION / CLASS-ACTION POSITION — TODO].
14. Contact
Numen Labs LLC, [REGISTERED ADDRESS]. General: [CONTACT EMAIL]. Privacy: [PRIVACY CONTACT EMAIL]. Security: [SECURITY CONTACT EMAIL]. Company registration number [COMPANY NUMBER]; VAT number [VAT NUMBER] if applicable.
These Terms, together with the Privacy Policy, are the entire agreement between us about the Service. If any provision is unenforceable, the rest stays in force. Our failure to enforce a provision is not a waiver of it.