Privacy Policy

Last updated: 29 October 2026 · Version 2.21 · The German version is legally binding.

Preliminary note: two different roles

This policy describes two things that must be kept strictly apart under data protection law.

• Part A covers data we decide about ourselves — visiting this website, contacting us, or administering your user account. Here we are the controller within the meaning of Art. 4(7) GDPR.
• Part B covers content and data a company brings into the platform as our customer — documents, messages, knowledge bases, workflows. That company alone decides about this data. We process it exclusively on its behalf and on its instructions. Here we are a processor pursuant to Art. 28 GDPR.

If you are an employee, customer or contact of a company that uses our platform, your rights are primarily directed at that company. We support it and forward any request without delay.

Part A — Where we are the controller

1. Controller

Beyondles UG (haftungsbeschränkt)
Bahnhofstraße 44, 14612 Falkensee, Germany
Represented by its Managing Director Niclas Krieg
Amtsgericht Potsdam, HRB 42428 P
Email: [email protected]
Further details are available in our Legal Notice.

2. Data protection officer

We are not required to appoint a data protection officer. The threshold of § 38 BDSG — at least twenty people constantly engaged in automated processing of personal data — is not met, and our core activity does not require large-scale regular monitoring of individuals. Please direct privacy enquiries to the address above.

3. Visiting this website

When you access our website, your browser automatically transmits information to our server which is temporarily stored in log files: IP address, date and time of access, page accessed, volume of data transferred, referring page, browser type and operating system. This data serves technical provision, stability and defence against attacks.

The legal basis is our legitimate interest in secure and functional operation (Art. 6(1)(f) GDPR). Log files are deleted after 30 days at the latest.

Fonts. All fonts used on this website are served from our own server. No connection is made to Google Fonts or any other external font service; your IP address is not transmitted there.

4. Contact, Founding Partner application, guide and Use Case Navigator

If you use our application or contact form, we process the data you provide: name, business email address, company, role, company website, and your notes on the task at hand and your experience with artificial intelligence.

We process this data solely to review your enquiry, contact you and prepare a possible collaboration. The legal basis is pre-contractual measures at your request (Art. 6(1)(b) GDPR) and our legitimate interest in handling business enquiries (Art. 6(1)(f) GDPR).

The data is stored in our database and additionally sent to us by email. If no collaboration results, we delete it as soon as it is no longer required for that purpose, at the latest after 24 months, unless statutory retention periods apply.

Guide and Use Case Navigator. On the public pages /leitfaden and /navigator you can request free material. The Use Case Navigator first takes you through a few questions about your situation and builds a recommendation from your answers.

We process your answers to those questions, any free text you choose to enter, your name, business email address, the company (with an explicit „prefer not to say“ option) and, optionally, a phone number so that we can clarify questions by phone rather than in writing; if you expressly ask for a conversation, it is required for that purpose. As proof of your consent we also record the time, IP address and browser identifier.

Delivery of the material and the follow-up email series are deliberately combined into one consent: the material is given in return for the sign-up, and the form says so. Consent takes effect only when you click the confirmation link in the first email (double opt-in); before that we send you nothing but that confirmation message and deliver neither a result nor any material. The legal basis is your consent (Art. 6(1)(a) GDPR, § 7(2) no. 2 UWG), which you can withdraw at any time with one click in every email.

A request for a conversation is separate, voluntary and not pre-selected. If you do not tick it, you still receive the material you asked for. The legal basis is handling your enquiry (Art. 6(1)(b) GDPR). Such a request is an enquiry, not a booked appointment.

Use of artificial intelligence. To phrase the Use Case Navigator's recommendation we send your answers and any free text to a model provider through our platform interface. Your contact details are not transmitted — the model sees the task, not the person. The selection of cases, all ratings and all source references come from our reviewed catalogue, not from the model. The providers we use are listed in section 11. If no model is reachable, you receive the same recommendation in a rule-based form.

Retention. If a request stays unconfirmed we delete it in full after 30 days at the latest. Confirmed contacts are stored until you withdraw consent or the purpose ceases, for no longer than 24 months. A stored assessment without a confirmed contact has its free text removed after the same period.

Session assignment. So that your assessment is assigned to you and not to someone else, the Use Case Navigator sets a strictly necessary cookie named nav_sid with a lifetime of 24 hours. It holds a random identifier, no personal data. No reach or advertising measurement takes place on these pages.

5. User account and sign-in

A user account is required to use the platform. We process your name, email address, a password stored only in encrypted form, your membership of an organisation, your role within that organisation, and the times of sign-in and last activity. The legal basis is performance of the usage contract (Art. 6(1)(b) GDPR).

For session management we set strictly necessary cookies, among them platform-auth-token on the .beyondles.ai domain. All cookies with their purpose and retention are listed in section 7.

Sign-in records. For each Beyondles account you sign in with in a browser, we store a sign-in record (session). It contains the IP address and the browser identifier (user agent) at the time of sign-in, the time of sign-in and of last use, the organisation concerned, whether the sign-in was confirmed with a second factor, and a check value of the platform-browser cookie, not the cookie itself. When the record ends, we note the time and the reason. This lets you use several accounts in one browser and switch between them, and signing out ends the sign-in of that account in this browser across all areas of the platform. The purposes are secure sign-in, switching between accounts and protection against misuse. The legal basis is performance of the usage contract (Art. 6(1)(b) GDPR) and our legitimate interest in the security of the platform (Art. 6(1)(f) GDPR).

A sign-in record ends when the account has not been used for 24 hours, and at the latest 30 days after the last sign-in with a password. It also ends when you sign out or change your password. We delete ended records 30 days after they end.

Record of acceptance. On registration we record when and in which version you accepted the General Terms and Conditions, this privacy policy and the data processing agreement. The legal basis is our statutory accountability obligation (Art. 5(2) GDPR) and our legitimate interest in documenting contract conclusions.

6. Billing and payments

To bill chargeable services we process invoicing data, contract data and usage figures. Payment processing is handled by Stripe; payment details such as card numbers are processed exclusively there and never reach our systems. The legal basis is performance of the contract (Art. 6(1)(b) GDPR) and compliance with commercial and tax obligations (Art. 6(1)(c) GDPR). We retain invoicing records for ten years.

7. Cookies

We use strictly necessary cookies only. There is no audience measurement and no tracking for advertising purposes. A consent banner is therefore not required.

The legal basis is § 25(2)(2) TDDDG and our legitimate interest in a functional service (Art. 6(1)(f) GDPR).

Part B — Where we process on your behalf

8. Our role

For all content a company brings into the platform as our customer, or causes to arise there, that company is the controller. We process such data exclusively on its instructions as a processor. The basis is the data processing agreement, which forms part of the contract between us and the respective company.

We do not use this content for our own purposes — in particular not to develop or train our own models, not for advertising, and not for any analysis beyond the assignment. That applies to us; which of the model providers we use exclude training on your data and which do not is stated expressly in section 11.

9. What data is processed

Nature and scope are determined solely by the customer. Typically this includes documents and files, messages and conversation histories, knowledge bases and their machine representations, tasks and workflows, data from connected third-party systems, and technical execution logs.

Data subjects may include employees, customers, prospects and business partners of the respective company. For connected services such as email, calendar, files, chats and meetings, this also includes people who communicate with the users, for example recipients of emails, participants in appointments and meetings or members of a chat, even if they do not use Beyondles themselves.

10. Where the data is stored

The platform runs entirely on servers of Hetzner Online GmbH in data centres in Germany (Nuremberg and Falkenstein). This expressly includes the knowledge and vector store and the knowledge graph database: both run on the same server and not with an external provider. Your content does not leave Germany for storage purposes.

A transfer occurs when a request is made to a language model — the content required to answer it is then transmitted to the respective model provider. This is described in section 11.

11. AI model providers

To provide AI-supported features we use third-party language models. Which provider is used for a given task is decided by an automatic router based on suitability and cost. Your organisation can define in its settings which providers and models are permitted.

Contractual position with the model providers. We use Anthropic, OpenAI, Google, Mistral and Microsoft (for the DeepSeek models) via their paid business interfaces. For all five, the use of your content for training is excluded by the terms of paid usage, no human review takes place, and a data processing agreement is in place with all five as part of the respective commercial terms.

Open models via Microsoft Foundry. Models by the developer DeepSeek are not obtained from DeepSeek itself. They run on Microsoft Foundry, Microsoft's AI service. Data is stored in the European Union (Sweden); individual requests may be processed in Microsoft data centres in the European Union or in the USA. Transfers to the USA are covered by the Standard Contractual Clauses and the EU-US Data Privacy Framework, to which Microsoft is certified. No content is transmitted to DeepSeek. This processing is governed by the Microsoft Product Terms, which exclude the use of your content for training and include a data processing agreement.

Decision model with upstream anonymisation. For small classification steps, such as which assistant handles a request, whether a message is urgent or whether an output touches a blocked topic, we use the decision model Jev by TypeSafe AI, Inc. (USA). Before any text is sent there, we replace personal data with placeholders on our own servers in the European Union, for example names, organisations, places, addresses, e-mail addresses, telephone numbers, bank details, identifiers and dates of birth. Names of organisations or web addresses may remain in individual cases where the classification needs them. TypeSafe also receives fixed configuration texts unchanged, such as the names and roles of assistants and the questions asked. TypeSafe returns a probability or a category, not text. The contract with TypeSafe includes a data processing agreement with EU standard contractual clauses and excludes training on the inputs. It does not provide a fixed deletion period for the transmitted text. Complete anonymisation cannot be guaranteed technically; a residual risk remains. We measure and document the detection rate regularly. If the anonymisation is unavailable, nothing is transmitted.

What you can control yourself. In your organisation's settings you define which providers and models are permitted for you. None of the providers listed here uses your content for training, and a data processing agreement is in place with each of them. With the five model providers named above, no human review takes place either; TypeSafe receives text only after anonymisation. For the web search (section 12), Linkup receives search queries only after anonymisation.

Change of providers. We reserve the right to change the models and providers used, provided the contractually owed functionality is preserved. We give at least 30 days' notice before adding a new sub-processor; until then an objection for good cause is possible. A change that would result in processing in a third country without an adequacy decision occurs only after prior explicit activation.

Connected Google services. When a user connects a Google account to Beyondles, the user signs in at Google directly. Each building block is its own connection. Depending on which building blocks are connected, this covers Google Calendar (reading the calendars the person can access, creating and changing events), Google Sheets and Google Drive (only files the person creates with Beyondles or picks themselves in a Google file picker, that is Docs, Slides, PDFs and the spreadsheets in them; reading and changing spreadsheets), Gmail (reading emails, creating drafts, labelling emails and sending on behalf of the person; only for accounts that Beyondles has enabled for this), Google Contacts, Google Tasks, the Google Workspace directory (name, email address and job title of colleagues) and Google Meet (meetings and transcripts). Beyondles then acts only on behalf of the person who signed in with Google and sees only what this person may open in Google themselves. Each building block requests only the permissions it needs. The data stays in the customer's Google account. We retrieve it only when a user, or an agent set up by that user, carries out a task.

The only thing we store permanently is the Google access token, encrypted. We delete it when the connection is removed. We keep a log of every retrieval with time, acting person, tool and result, but without content. We delete this log after 180 days. Content that an agent reads can become part of the conversation history and can be sent to a provider from this section to produce the answer. Section 14 applies to it.

We do not use this data to train AI models, do not sell it, do not use it for advertising and do not pass it on to third parties except as described in this policy. No human at Beyondles reads this data unless the user has given consent, it is necessary for security, the law requires it, or the data is aggregated and anonymised. Beyondles' use and transfer to any other app of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements. Google is not our sub-processor for this connection. Google processes the data on behalf of the customer under the customer's own contract with Google.

With whom we share Google user data. We share data received from Google APIs only with the following recipients, and only as far as needed for the feature the user is using:

• AI model providers (section 11: Anthropic, OpenAI, Google Gemini API, Mistral, Microsoft Foundry): the content an agent needs to produce the answer, and only to providers the organisation has permitted. All of them contractually exclude training on this data.
• TypeSafe AI, Inc. (decision model, section 11) and Linkup Technologies SAS (web search, section 12): only text in which personal data has first been replaced by placeholders on our own servers.
• Hetzner Online GmbH (Germany): operation of our servers, where the data is processed and the encrypted access token is stored.
• Cloudflare, Inc.: encrypted transport between the user, Beyondles and Google.
• Services the user chooses: if a user asks an agent to send Google data to another connected service (for example an email or a message in Slack or Microsoft Teams), exactly this content goes there.
• Authorities: only where a law or a court order obliges us to.

All processors are bound by contract pursuant to Art. 28 GDPR (section 12). Beyond this we do not share, transfer or disclose Google user data to anyone, including advertising networks, data brokers or information resellers. Our external development contractors have no access to Google user data.

Connected social media accounts (X, Instagram, LinkedIn). When a user connects an account at X (X Corp., USA), Instagram (Meta Platforms Ireland Ltd.) or LinkedIn (LinkedIn Ireland Unlimited Company) to Beyondles, we retrieve via that network's interface the data needed for the features the user has enabled. This can include profile details and the email address of the connected account, own and other people's posts, replies, followers, accounts followed, lists, likes, bookmarks and direct messages. Direct messages and replies also contain data of people who are not Beyondles customers themselves. For Instagram this additionally includes the comments on the customer's own posts, including the name, username and text of the people who comment, as well as statistics about posts and the account (for example reach and engagement). For LinkedIn this additionally includes the comments and reactions on the posts of a connected company page, including the name and text of the people who comment, the aggregated follower statistics of the page and, for a connected private account, the numbers of the member's own posts and the member's follower and connection count. For a connected company page we also read the page's notifications (mentions, comments and shares of the last 60 days, with the identifier of the post concerned and the time) and the statistics of page views by section and time range. We reply to, hide or delete comments only after the user approves it. We process this data on behalf of the customer (Art. 28 GDPR); the customer's legal basis is usually its legitimate interest in managing its own account (Art. 6(1)(f) GDPR). The data is used only for the task the user started or approved and is transmitted to the model providers named in section 11 for that purpose. It is not used to train AI models and is not sold. Every action that changes something on the network (post, like, follow, direct message) happens only after the user approves it. We store the access credentials for the network encrypted; they are deleted when the connection is removed. The networks themselves are independent controllers towards the user, and their own privacy notices apply in addition.

LinkedIn network and messages (data portability). When a member connects "LinkedIn network and messages", we use LinkedIn's Member Data Portability API (third party access). This interface is available only to members in the EU. The provider is LinkedIn Ireland Unlimited Company. At the member's request we read their own profile, posts, comments, reactions, connections, connection requests, synced contacts and messages. Also at the member's request we promptly read the member's own recent activity from LinkedIn's change log (Member Changelog): messages sent and received, connection requests, own comments, likes and reactions, and own posts. LinkedIn keeps this activity for about 28 days. The connection service does not store it, and it is visible only to the member. We only read and change nothing through this interface. The data is visible only to the member who made the connection, not to their employer, not to the administrators of their organisation and not to other users. We use it only for the tasks the member starts and pass it only for that task to the model providers named in this section. It is not used to train AI models, not sold and not used for advertising. Messages and contacts also contain data of third parties. We process the data on behalf of the customer (Art. 28 GDPR), as described in the previous paragraph. LinkedIn grants access for up to one year. We delete the data when the connection is removed. Export and deletion are possible at any time on request (sections 14 and 17).

LinkedIn Ads. When a customer connects LinkedIn Ads, we read its ad accounts, campaigns and reports. After the customer approves, we create or change campaigns and budgets, as well as a campaign's name, schedule and targeting and the status of an ad. To choose the targeting we look up attributes such as industry, job function, seniority and location and let LinkedIn estimate the approximate audience size. For audiences (Matched Audiences) the customer can upload company lists with company names and domains, and contact lists, and later add to, shrink or delete them; we read the status and match rate of the lists. Email addresses in contact lists are hashed with SHA-256 (turned into a non-reversible checksum) before they are sent to LinkedIn. The customer is the controller for these lists and confirms that it has a legal basis for using them. LinkedIn is an independent controller for its own processing for advertising purposes; LinkedIn's own terms and privacy notices apply. From LinkedIn's Ad Library we retrieve only public data.

How to delete data from connected social media accounts. You have several options. (a) Disconnect the account in the Suite under "Connections". The access credentials for the network are deleted immediately, and we stop retrieving data. (b) Remove the app "Beyondles" in the network's settings: on Instagram and Facebook (Meta) under "Apps and websites", on X and LinkedIn in their settings for connected apps. (c) Delete your Beyondles account or organisation. All data is then deleted as described in section 14. (d) Write to us at [email protected]. We handle your deletion request within 30 days. Content that has already been published on the network stays there. You delete it in the network itself.

Connected Microsoft services. When a user connects a Microsoft account to Beyondles, the user signs in at Microsoft directly. Depending on which building blocks are connected, this covers Outlook Mail (emails, attachments, drafts and sending on behalf of the person), Outlook Calendar (including calendars shared by colleagues), OneDrive and SharePoint (files, folders and sites), Teams meetings with transcripts and recordings, Teams chats and channels, contacts, tasks in To Do and Planner, OneNote, Bookings and the company directory. Beyondles then acts only on behalf of this person and sees only what this person may open in Microsoft 365 themselves. Each building block requests only the permissions it needs. The data stays in the customer's Microsoft 365. We retrieve it only when a user, or an agent set up by that user, carries out a task.

The only thing we store permanently is the Microsoft access token, encrypted. We delete it when the connection is removed. We keep a log of every retrieval with time, acting person, tool and result, but without content. We delete this log after 180 days. Content that an agent reads can become part of the conversation history and can be sent to a provider from this section to produce the answer. Section 14 applies to it.

We do not use this data to train AI models and do not pass it on to third parties except as described in this policy. Microsoft is not our sub-processor for this connection. Microsoft processes the data on behalf of the customer under the customer's own contract with Microsoft.

Further connected services (Trello, GitHub, LinkedIn company pages, MCP servers). Trello (Atlassian): boards, lists, cards, comments, attachments and board members the connected person has access to, including creating and changing cards. GitHub (GitHub, Inc., USA): profile of the connected account, repositories, files and code, issues, pull requests and comments. These often contain names and user names of other contributors. LinkedIn company pages: posts and statistics of a page the connected person may manage, and publishing on behalf of the page. The paragraph on social media accounts applies. MCP servers: a customer can enter the address of a server that offers tools under the Model Context Protocol. When a user or agent calls such a tool, we send the inputs of the call to that server and receive its answer. The customer chooses the server and its operator. Beyondles knows neither its location nor how it handles the data. The customer checks whether it may transfer data there and makes the necessary agreements with the operator.

Apollo (research of business contacts). When a customer connects Apollo (Apollo.io, provider ZenLeads, Inc., USA) with its own API key, we search Apollo's database for people and companies and enrich them. We read business contact data of third persons: name, title, company, LinkedIn address and business e-mail address, plus company details. We never request private e-mail addresses or phone numbers. We write nothing in Apollo. The purpose is finding contact persons for the customer's organisation (lead research). The customer's legal basis is as a rule its legitimate interest in approaching business contacts (Art. 6(1)(f) GDPR, B2B). The obligation to inform the persons concerned (Art. 14 GDPR) lies with the customer as controller. We store the key encrypted and delete it when the connection is removed. The customer can revoke the key only in Apollo itself (Settings, Integrations, API Keys). We delete the retrieval log without content after 180 days. Apollo states in its privacy policy that ZenLeads, Inc. is certified under the EU-US Data Privacy Framework. Where this certification does not apply, transfers to the USA rely on standard contractual clauses between the customer and Apollo. Apollo is not our sub-processor for this connection. The customer pays the credits directly to Apollo.

GitHub app “Beyondles GitHub”. Instead of a personal access token, a customer can install our GitHub app in its GitHub account or organisation. The customer chooses which repositories the app can see. We store only the installation ID, the account and the repository selection. For each retrieval we obtain an access token from GitHub that expires after one hour at most, and we do not store it permanently. The app can read repositories, files and code, issues, pull requests, reviews, comments, check results and organisation members. On the customer's behalf it can write: create branches and commits, open and comment on pull requests and issues. Such contributions appear at GitHub under the app's name. Merging, deleting repositories, changing workflows and access to secrets are excluded. GitHub notifies us of events in the selected repositories, such as new pull requests, reviews, pushes, issues and check results. Of these notifications we keep only type, time, repository and installation, and delete them after 180 days. If the customer uninstalls the app, access ends immediately and we remove the connection.

Work data of employees. Data from GitHub, Trello and similar services often shows who worked on what. If a customer uses it to evaluate the work of individual employees, for example in a metrics Lab, it processes employee data. The legal basis, transparency towards employees and any required involvement of the works council (Section 87(1) no. 6 German Works Constitution Act) are the customer's responsibility as employer. We do not evaluate this data for our own purposes and make no decisions about employees (see section 16).

For all connected services: we process the data on behalf of the customer (Art. 28 GDPR) and to perform the contract with the customer (Art. 6(1)(b) GDPR). The customer's own legal basis depends on its purpose. We retrieve data only when a user, or an agent set up by that user, carries out a task, and we do not use it to train AI models. We store access tokens and credentials encrypted and delete them when the connection is removed. For every service we keep a log of every retrieval with time, acting person, tool and result, but without the inputs and without the content of the answer. We delete this log after 180 days. Section 14 applies to the AI agent logs. The providers Google, Microsoft, X, Meta, LinkedIn, Atlassian, GitHub and Apollo are established in the USA or belong to groups established there. They are not our sub-processors for these connections. Data goes to them only when the customer causes this through the connection. Where a provider is certified under the EU-US Data Privacy Framework, a transfer to the USA is based on the European Commission's adequacy decision, otherwise on the safeguards the provider has agreed with the customer, in particular Standard Contractual Clauses. The provider's own privacy notices apply in addition.

Switching off individual functions. For every connection, each single function can be switched off under "Manage functions", for example reading direct messages, emails or files. The person who created the connection may switch functions off. For connections shared with a collection its members may do so as well, for connections for the whole organisation also an admin. A function that is switched off cannot be used through this connection by any agent, Lab or other application. An attempt is refused and logged without content. When connecting, we point out if a connection can read private content. This lets the customer limit the processing to what is necessary (data protection by design and by default, Art. 25 GDPR).

12. Further sub-processors

Beyond the model providers we use the following service providers. All are bound by contract pursuant to Art. 28 GDPR.

The external development contractors have no access to the production databases and no access to credentials. However, for debugging purposes they can view operational logs of the running system, which may contain personal data. They are independent individuals; each of them is bound separately by a data processing agreement, a confidentiality agreement and the EU Standard Contractual Clauses. On request we will provide customers with their names and addresses.

Web search with upstream anonymisation. When an agent in Beyondles HorAIzon searches the internet, we send the search query to Linkup Technologies SAS (Fresnes, France). Linkup returns public search results, that is titles, web addresses and text snippets. Before a search query leaves our platform, it passes through the same anonymisation on our own servers in the European Union as for the decision model (section 11). Names of persons, e-mail addresses, telephone numbers, postal addresses, bank details, identifiers and dates of birth are replaced by placeholders. Company names, place names and web addresses deliberately stay in the search query, because otherwise a search for a company or a place finds nothing. If the anonymisation is unavailable, no search is sent. Linkup operates its systems in the European Union. Some of Linkup’s sub-processors are established in the USA; transfers to them are covered by the EU Standard Contractual Clauses. A data processing agreement under Art. 28 GDPR is in place with Linkup. Linkup processes the search queries only to provide the search and the related support. No use for its own purposes is agreed. Under the agreement, Linkup may retain queries and results for the duration of the contract until we instruct deletion. The agreement provides no automatic deletion period for search queries; after our instruction Linkup deletes within 30 days. Technical logs are deleted after 14 days, backups after 30 days.

13. Transfers to third countries

Where data is processed outside the European Economic Area, we rely on the EU Commission's Standard Contractual Clauses pursuant to Implementing Decision (EU) 2021/914 and, where the respective provider is certified, additionally on the EU-US Data Privacy Framework. For the People's Republic of China none of these safeguards exists; transfers there occur exclusively following explicit activation by the responsible organisation (see section 11).

14. Retention and deletion

A customer's content is stored for the term of the contract. After the contract ends we make the data available for export in a common, machine-readable format for 30 days.

It is then deleted from the main databases. Copies in the knowledge and vector store and in the knowledge graph database are removed as part of regular clean-up, at the latest within 90 days after the end of the contract. Backup copies expire after no more than 30 days and are not separately purged.

AI agent logs. For every tool call by an agent we log which tool ran when, whether it succeeded and how long it took. For tools of connected services (for example X, Instagram, LinkedIn or email) and of the Labs we store neither the inputs nor the content of the response. These logs are deleted automatically after 90 days. Larger intermediate results an agent needs during a running task are deleted after 24 hours, or at the end of a longer-running task.

Independently of this, we delete or block data on the customer's instruction. Statutory retention obligations — in particular commercial and tax law periods of up to ten years for invoicing records — remain unaffected.

15. Security of processing

We implement technical and organisational measures pursuant to Art. 32 GDPR. These include encryption in transit, encrypted storage of credentials for third-party systems, separation of data per customer at database, vector and graph level, separate database roles for application, schema change and maintenance, access protection for administrative interfaces, network filtering, logging of security-relevant events, regular backups with tested restoration and strictly restricted access, and inspection rules to detect impermissible inputs and outputs.

The full description is Annex 2 to the data processing agreement.

16. Automated decisions and the AI Act

The service produces results with the help of statistical models. These may be factually wrong or incomplete even when they appear convincingly worded. We do not take automated decisions within the meaning of Art. 22 GDPR — that is, decisions producing legal effects concerning a person without human involvement. If a customer configures workflows such that results are adopted without review, that customer is responsible for it; we provide approval steps, guardrails and logs for this purpose.

Part C — Common provisions

17. Your rights

Under the GDPR you have the right to access (Art. 15), rectification (Art. 16), erasure (Art. 17), restriction of processing (Art. 18), data portability (Art. 20) and objection (Art. 21), as well as the right to withdraw consent with future effect (Art. 7(3)).

To exercise your rights, an informal message to [email protected] is sufficient. If your request concerns content processed by a company via our platform, we forward it to that company without delay and support it in responding.

18. Right to lodge a complaint

You have the right to lodge a complaint with a data protection supervisory authority (Art. 77 GDPR). The authority responsible for us is:

Die Landesbeauftragte für den Datenschutz und für das Recht auf Akteneinsicht Brandenburg (LDA Brandenburg), Stahnsdorfer Damm 77, 14532 Kleinmachnow, Germany.

19. Changes to this privacy policy

We adapt this policy whenever changes to our processing make it necessary. The version published on this page applies. We notify existing customers of material changes — in particular the addition of new sub-processors — by email at least 30 days in advance.