Terms of Service

B2B only — not available to consumers

§ 1 Scope

These General Terms and Conditions (GTC) govern all contracts between NMA Venture Capital GmbH ("Provider") and the customer ("Customer") for the Plietsch Platform. The offer is exclusively aimed at entrepreneurs (§ 14 BGB), legal entities under public law, or special funds. Contracts with consumers (§ 13 BGB) are strictly excluded.

§ 2 Contract Conclusion

The contract is concluded when the Customer accepts these Terms by checkbox and completes registration, and the Provider activates access or confirms setup by email.

§ 3 Services & Availability

The Provider supplies the SaaS software via the Internet as described in the Service Description. Service availability is guaranteed at 99.0% annual average, excluding scheduled maintenance windows and third-party cloud node outages. The Provider may develop and iterate the software provided that core functionality remains unaffected.

§ 4 Customer Obligations

The Customer must provide truthful registration details, keep credentials secure, and refrain from placing unlawful task configurations, malware, or violating security perimeters of targeted hosts.

§ 5 Remuneration & Payment

All prices are net of statutory VAT. Subscription payment is handled in partnership with Mollie B.V. (Netherlands). Usage fees are due in advance at the start of each billing period. Provider may suspend access after written notice if payment falls into default.

§ 6 Term & Termination

Monthly plan: Indefinite; terminable at the end of the current month. Annual plan: 12-month minimum; auto-renews for another 12-month period unless terminated with 1 month notice prior to the end of the term. Termination can be performed via account settings or email.

§ 7 Liability

Unlimited liability for intent, gross negligence, and personal injury. For slight negligence: liability only for breach of material contractual obligations (cardinal obligations), limited to typical foreseeable damage. Indirect damages and lost profits are excluded.

§ 8 Data Protection; Derived Labels and Anonymized Structural Statistics

8.1 Data protection

Both parties comply with GDPR. Where the Provider processes personal data on behalf of the Customer (e.g. executing agent tasks), a separate Data Processing Agreement (DPA) is executed, forming part of this contract.

8.2 Definitions

"Derived labels" are records of how the Customer's users respond to a decision the Plietsch classifier has made: a user's correction of a triage decision, a one-click disagreement, and passive outcome signals such as whether and how quickly a reply was sent. A derived label consists of the decision identifier, the class proposed by Plietsch, the class chosen by the user or the outcome observed, timestamps and the Customer's tenant identifier. "Anonymized structural statistics" are aggregate figures computed inside the Customer's tenant, such as thread length, number of participants, reply latency and the share of decisions per class, that leave the tenant only in aggregated form and cannot be traced to a message, a person or the Customer. Neither category contains message text, subject lines, attachments, contact details or any other content of the Customer's messages or documents; such content never leaves the Customer's tenant for this purpose.

8.3 Purpose and permission

The Customer permits the Provider to use derived labels and anonymized structural statistics for the Provider's own purpose of improving the classification models that serve all Plietsch customers, in particular a shared classification model trained on pooled derived labels, and for calibrating those models to the Customer's tenant. The Provider acts as controller for this use; it is not processing on the Customer's instruction under the DPA. Insofar as derived labels are personal data, the Provider bases this use on Art. 6(1)(f) GDPR (legitimate interest: improving a product that all customers, including the Customer, benefit from; the data is limited to decision-level signals without message content; the persons affected are the Customer's users acting in a professional context; the opt-out under 8.6 is available at any time). Anonymized structural statistics are not personal data.

8.4 Tenant partitioning and erasure

The Provider keeps the corpus of derived labels partitioned per Customer and re-derivable, so that pooled training sets and models are always rebuilt from the partitioned corpus and the corpus is never folded irreversibly into model weights. On termination of the contract, on opt-out under 8.6, or on an erasure request, the Provider deletes the Customer's partition within 30 days and rebuilds the pooled training set and the shared models without the Customer's contribution at the next scheduled re-derivation, at the latest within 90 days. Anonymized structural statistics that have already been aggregated cannot be attributed to the Customer and are not withdrawn.

8.5 No sale, no third parties, EU hosting

Derived labels and anonymized structural statistics are stored and processed exclusively on the Provider's servers in the European Union, are never sold, never passed to third parties and never used for advertising. Models are trained on EU infrastructure and exclusively on open-weight base models; no derived label is sent to an external model provider.

8.6 Opt-out

The Customer may opt out of the use under 8.3 at any time with effect for the future, in text form (an email to info@nma.vc suffices). After opt-out, no further derived labels or statistics leave the Customer's tenant, and 8.4 applies to the existing contribution. Opt-out does not restrict the Customer's use of the Plietsch Platform: the Customer continues to receive the shared models, and calibration on the Customer's own corrections inside its tenant continues as part of the service (Art. 6(1)(b) GDPR).

8.7 Google-sourced data

For messages and documents read from a Google account (e.g. Gmail, Google Workspace) via a connected source, the Provider complies with the Google API Services User Data Policy, including the Limited Use requirements. No derived label and no structural statistic derived from Google-sourced data is added to the pooled corpus or used to train the shared models; such signals are used only inside the Customer's own tenant to calibrate the models for the Customer.

§ 9 Governing Law & Jurisdiction

German law applies, excluding the UN CISG. Exclusive place of jurisdiction for merchant Customers: Hamburg, Germany.