Terms & Conditions.

Eigenwelt Plus is the hosted inference and team platform operated by Poensgen Technology UG (haftungsbeschränkt), Berlin, Germany (“Eigenwelt”, “we”, “us”). These Terms govern the platform at platform.eigenweltlabs.com, paid Eigenwelt model inference, and connected team, Knowledge Hub, intake, and task features (the “Service”).

The Service is for businesses and professionals. If you accept these Terms for a law firm or other organisation (“Customer”), you confirm that you may bind it. The free LegalWork desktop App has separate terms.

The service

Eigenwelt Plus provides organisation accounts, paid model inference, premium features, member and settings management, usage reporting, a Knowledge Hub for supported workflows, integrations, plugins, MCP servers, and presets, and intake, which turns incoming emails and API submissions into assigned tasks. Custom deployments, customer-specific models, and continual-learning work may use a separate order form.

There are two plans, Plus and Pro. Both include the Eigenwelt models with the included usage described below, plus the Knowledge Hub and intake; they differ in price and in the amount of included AI usage. The plan, price, and scope are those shown in the order and on the billing page.

Model inference, regions, and ZDR

Eigenwelt's gateway does not store prompt or output content and retains only content-free metering data such as model, token counts, cost, status, and timestamps. We do not use paid inference content to train shared or public models.

Every model carries a hosting region, shown in the platform and in LegalWork. The European Union is the default: unless an administrator has enabled a further region, paid inference for Customer runs on model infrastructure in the EU only. A model we add later — including one on a new provider — is off until an administrator turns it on.

An administrator can enable further regions, currently the United States. That selection applies to all of Customer's users and is a documented instruction to send inference content to model infrastructure in the selected region. Such transfers are covered by an adequacy decision or appropriate safeguards under Chapter V GDPR. Customer is responsible for assessing whether a region is consistent with its confidentiality, professional, and data-protection duties.

Zero data retention (“ZDR”) is enabled for Eigenwelt’s accounts with our inference subprocessors; prompt or output content is not used to train models. One qualification applies: a provider's published policy may permit temporary access and retention for up to two weeks when investigating abnormal errors or suspected malicious activity. ZDR does not apply to account, billing, security, and usage records or to Knowledge Hub items Customer intentionally stores. Per-provider details and the list of providers are in the Eigenwelt Plus Privacy Notice.

ZDR also applies to the Gemini models served through Google Cloud (Vertex AI): Google has switched off abuse-monitoring logging for Eigenwelt, so prompts sent to these models are not logged or retained for abuse monitoring either.

ZDR also applies to the Claude and GPT models: we obtain them through OpenRouter only from zero-data-retention endpoints at Microsoft Azure, Amazon Web Services, and Google Cloud; requests are not sent to Anthropic's or OpenAI's own interfaces.

Eigenwelt Europe EigenJev is operated by Eigenwelt on Google Cloud in the Netherlands (europe-west4). Eigenwelt US Jev is provided by TypeSafe AI, Inc. in the United States, with ZDR enabled for Eigenwelt’s TypeSafe account. An administrator must enable both the US region and the model to use Eigenwelt US Jev.

Accounts and administrators

Each account and seat is for one named user. Users must keep credentials secure and may not share accounts, tokens, or API keys. Customer is responsible for its users and for removing access when no longer authorised.

Administrators may manage users, seats, billing, usage settings, and shared Hub items, and decide which models and hosting regions the firm uses. Those choices apply to all of Customer's users. Customer is responsible for appointing administrators and obtaining required notices and consents. Administrator actions bind Customer.

Plans, seats, and payment

Paid plans are recurring and billed per seat. The plan, price, billing interval, seat count, included usage, and features shown at checkout, in the dashboard, or in an order form are part of this agreement. Prices exclude VAT and other taxes unless stated otherwise; VAT is calculated and shown at checkout.

Each plan is offered on two billing intervals. On yearly billing the term is one year, the fee is due in advance for the year, and the subscription renews for a further year unless cancelled before renewal. On monthly billing the term is one month and renews monthly. The billing interval is chosen at checkout and can later be changed through the payment provider's customer portal where offered; a change takes effect at the next renewal.

Self-service subscriptions renew until an administrator cancels. Cancellation and plan downgrades take effect at the next renewal. Seat additions and upgrades may take effect immediately with charges prorated over the remaining term; on yearly billing an added seat is prorated to the end of the current year. Seat reductions cannot go below current members and pending invitations and do not create a mid-period credit. Paid fees are non-refundable except where required by law or stated at purchase.

Free trial

A firm's first subscription, on either plan, starts with a free trial of seven days. A payment method is collected at checkout but is not charged during the trial. Unless it is cancelled before the trial ends, the trial converts automatically into the paid subscription chosen at checkout and the first payment falls due. The trial can be cancelled at any time in the dashboard under “Billing”.

At most one trial is granted per firm; a prior subscription disqualifies. We notify Customer in text form before the trial ends.

Included AI usage

Each paid seat includes the inference allowance shown for its plan. The allowance is quoted as a monthly amount and granted weekly: the quoted monthly amount is spread across the weeks of the year (monthly amount × 12 ÷ 52) and made available to each seat per calendar week. The week follows ISO 8601 and begins on Monday at 00:00 UTC. Usage depends on the model and input, cached-input, and output tokens processed. Allowances are assigned per user, reset weekly, do not roll over into the following week, have no cash value, and cannot be transferred or resold.

Usage is counted against the allowance, and against any purchased credits, at the model prices in effect when the request is made. Model prices are shown per model for input, cached-input, and output tokens on the platform's Models page. They consist of the price our provider charges for the model plus a platform fee, and follow changes in provider prices, exchange rates, and the platform fee. We may change model prices with immediate effect for future requests; usage already processed is not repriced. Model prices are not part of the plan, and changing them is not a plan change under “Changes, suspension, and termination”.

Requests pause when a user exhausts the weekly allowance and resume when the following week begins. The app shows usage consumed and when the allowance next resets. If optional additional usage is offered and enabled, included usage is consumed first and further usage draws from purchased credits at displayed rates and may be subject to a separate monthly cap. Administrators may be able to set spending caps. Credits are non-transferable and non-refundable except where law or purchase terms say otherwise. Metering may slightly overshoot a limit because final usage is known only after a request completes.

Knowledge Hub and Customer content

Hub items are shared with authorised members according to roles and permissions. Customer is responsible for what its users publish and must have the right to share it. This includes items the assistant creates or writes into the shared library on a user's instruction, which count as published by that user. Credentials may be included only through an expressly protected sharing feature. Automated secret detection does not replace user review.

Customer retains its rights in prompts, Hub items, configurations, and other content. To the extent permitted by law, Customer owns generated output. Customer grants us and our service providers only the rights needed to provide, secure, and support the Service and comply with law.

Intake and tasks

Intake is off until an administrator turns it on. Customer decides which senders are admitted, who sees tasks, and which model intake uses, and is responsible for informing senders about the processing of their messages where required. Emails sent to intake addresses are received by our provider Brevo; details are in the Eigenwelt Plus Privacy Notice.

AI processing for intake has its own budget for the firm, shown in the intake settings. Once it is reached, new messages still become tasks, but without AI extraction and assignment, until the budget resets. Tasks and assignments created by AI must be reviewed like any other AI output.

Confidentiality

Each party will protect the other's confidential information with reasonable care, use it only for this agreement, and disclose it only to personnel and providers who need it and are bound by confidentiality, or where law requires disclosure. Customer Content is Customer's confidential information.

Acceptable use and professional responsibility

Customer must use the Service lawfully and in line with professional, confidentiality, privilege, supervision, and data-protection duties. Users must not bypass seat, usage, safety, or security controls, interfere with the Service, or violate another person's rights.

Eigenwelt Plus assists with legal work but does not provide legal advice. AI output can be incorrect, incomplete, or outdated. A qualified professional must review and verify output and citations before relying on or sharing them.

Changes, suspension, and termination

We may update models and features or apply reasonable safety, security, rate, and capacity controls. We may suspend access for a security risk, unlawful use, breach, non-payment, or provider outage and will use reasonable efforts to give notice and restore access where possible.

Material adverse plan changes receive at least 30 days' notice and apply no earlier than the next renewal unless required sooner by law or urgent security needs. The EU-by-default hosting of paid inference and ZDR as described above are core paid commitments; we will not change them to Customer's detriment without notice under this paragraph. When a subscription ends, hosted inference and premium features stop. Customer should export Hub data it needs before termination.

Warranty and liability

Except for express commitments above, the Service is provided “as is” and “as available”. We do not warrant uninterrupted service or accurate output. Nothing limits liability for intent, gross negligence, injury to life, body or health, an expressly assumed guarantee, or mandatory liability. For slight negligence, we are liable only for breach of a material contractual obligation and foreseeable damage typical of this agreement.

Changes to terms, law, and contact

We may update these Terms prospectively with reasonable notice of material changes. German law applies, and the courts of Berlin have jurisdiction to the extent permitted by law.

Questions: chris@eigenweltlabs.com or eigenweltlabs.com/contact.